Tootfinder

OpenAI and Apollo Research trained o3 and o4-mini versions to not engage in "scheming", or secretly pursuing undesirable goals, reducing "covert actions" ~30X (Radhika Rajkumar/ZDNET)
https://www.zdnet.com/article/ai-models-kn

AI models know when they're being tested - and change their behavior, research shows
OpenAI and Apollo Research tried to stop models from lying - and discovered something else altogether.

https://www.axios.com/2025/12/16/ai-models-hacking-stanford-openai-warnings
Leaders from Amazon and Anthropic are testifying before two house committees today.

AI models are perfecting their hacking skills
This is the worst AI tools will likely ever perform, and they're already unnerving researchers and developers.

A Study on Thinking Patterns of Large Reasoning Models in Code Generation
Kevin Halim, Sin G. Teo, Ruitao Feng, Zhenpeng Chen, Yang Gu, Chong Wang, Yang Liu
https://arxiv.org/abs/2509.13758

A Study on Thinking Patterns of Large Reasoning Models in Code Generation
Currently, many large language models (LLMs) are utilized for software engineering tasks such as code generation. The emergence of more advanced models known as large reasoning models (LRMs), such as OpenAI's o3, DeepSeek R1, and Qwen3. They have demonstrated the capability of performing multi-step reasoning. Despite the advancement in LRMs, little attention has been paid to systematically analyzing the reasoning patterns these models exhibit and how such patterns influence the generated code. …

Can Current AI Models Count What We Mean, Not What They See? A Benchmark and Systematic Evaluation
Gia Khanh Nguyen, Yifeng Huang, Minh Hoai
https://arxiv.org/abs/2509.13939 htt…

Can Current AI Models Count What We Mean, Not What They See? A Benchmark and Systematic Evaluation
Visual counting is a fundamental yet challenging task, especially when users need to count objects of a specific type in complex scenes. While recent models, including class-agnostic counting models and large vision-language models (VLMs), show promise in counting tasks, their ability to perform fine-grained, intent-driven counting remains unclear. In this paper, we introduce PairTally, a benchmark dataset specifically designed to evaluate fine-grained visual counting. Each of the 681 high-reso…

💔 Are Foundation Models Ready for Industrial Defect Recognition? A Reality Check on Real-World Data
#computing

Are Foundation Models Ready for Industrial Defect Recognition? A Reality Check on Real-World Data
Foundation Models (FMs) have shown impressive performance on various text and image processing tasks. They can generalize across domains and datasets in a zero-shot setting. This could make them suitable for automated quality inspection during series manufacturing, where various types of images are being evaluated for many different products. Replacing tedious labeling tasks with a simple text prompt to describe anomalies and utilizing the same models across many products would save significant…

Defending Diffusion Models Against Membership Inference Attacks via Higher-Order Langevin Dynamics
Benjamin Sterling, Yousef El-Laham, M\'onica F. Bugallo
https://arxiv.org/abs/2509.14225

Defending Diffusion Models Against Membership Inference Attacks via Higher-Order Langevin Dynamics
Recent advances in generative artificial intelligence applications have raised new data security concerns. This paper focuses on defending diffusion models against membership inference attacks. This type of attack occurs when the attacker can determine if a certain data point was used to train the model. Although diffusion models are intrinsically more resistant to membership inference attacks than other generative models, they are still susceptible. The defense proposed here utilizes criticall…

Evaluation Awareness Scales Predictably in Open-Weights Large Language Models
Maheep Chaudhary, Ian Su, Nikhil Hooda, Nishith Shankar, Julia Tan, Kevin Zhu, Ashwinee Panda, Ryan Lagasse, Vasu Sharma
https://arxiv.org/abs/2509.13333

Evaluation Awareness Scales Predictably in Open-Weights Large Language Models
Large language models (LLMs) can internally distinguish between evaluation and deployment contexts, a behaviour known as \emph{evaluation awareness}. This undermines AI safety evaluations, as models may conceal dangerous capabilities during testing. Prior work demonstrated this in a single $70$B model, but the scaling relationship across model sizes remains unknown. We investigate evaluation awareness across $15$ models scaling from $0.27$B to $70$B parameters from four families using linear pr…

Adobe updates Firefly's app to add a prompt-based video editor using Runway's Aleph model, and adds image and video models from Black Forest Labs and others (Ivan Mehta/TechCrunch)
https://techcrunch.com/2025/12/16/adob

Adobe Firefly now supports prompt-based video editing, adds more third-party models | TechCrunch
Adobe is updating its AI video-generation app, Firefly, with a new video editor that supports precise prompt-based edits, as well as adding new third-party models for image and video generation, including Black Forest Labs' FLUX.2 and Topaz Astra.

Can Large Language Models Robustly Perform Natural Language Inference for Japanese Comparatives?
Yosuke Mikami, Daiki Matsuoka, Hitomi Yanaka
https://arxiv.org/abs/2509.13695 ht…

Can Large Language Models Robustly Perform Natural Language Inference for Japanese Comparatives?
Large Language Models (LLMs) perform remarkably well in Natural Language Inference (NLI). However, NLI involving numerical and logical expressions remains challenging. Comparatives are a key linguistic phenomenon related to such inference, but the robustness of LLMs in handling them, especially in languages that are not dominant in the models' training data, such as Japanese, has not been sufficiently explored. To address this gap, we construct a Japanese NLI dataset that focuses on comparative…

All Models Are Wrong, But Can They Be Useful? Lessons from COVID-19 Agent-Based Models: A Systematic Review
Emma Von Hoene, Sara Von Hoene, Szandra Peter, Ethan Hopson, Emily Csizmadia, Faith Fenyk, Kai Barner, Timothy Leslie, Hamdi Kavak, Andreas Zufle, Amira Roess, Taylor Anderson
https://arxiv.org/abs/2509.13346

All Models Are Wrong, But Can They Be Useful? Lessons from COVID-19 Agent-Based Models: A Systematic Review
The COVID-19 pandemic prompted a surge in computational models to simulate disease dynamics and guide interventions. Agent-based models (ABMs) are well-suited to capture population and environmental heterogeneity, but their rapid deployment raised questions about utility for health policy. We systematically reviewed 536 COVID-19 ABM studies published from January 2020 to December 2023, retrieved from Web of Science, PubMed, and Wiley on January 30, 2024. Studies were included if they used ABMs …

Dual-Actor Fine-Tuning of VLA Models: A Talk-and-Tweak Human-in-the-Loop Approach
Piaopiao Jin, Qi Wang, Guokang Sun, Ziwen Cai, Pinjia He, Yangwei You
https://arxiv.org/abs/2509.13774

Dual-Actor Fine-Tuning of VLA Models: A Talk-and-Tweak Human-in-the-Loop Approach
Vision-language-action (VLA) models demonstrate strong generalization in robotic manipulation but face challenges in complex, real-world tasks. While supervised fine-tuning with demonstrations is constrained by data quality, reinforcement learning (RL) offers a promising alternative. We propose a human-in-the-loop dual-actor fine-tuning framework grounded in RL. The framework integrates a primary actor for robust multi-task performance with a refinement actor for latent-space adaptation. Beyond…

Comprehensive Evaluation of CNN-Based Audio Tagging Models on Resource-Constrained Devices
Jordi Grau-Haro, Ruben Ribes-Serrano, Javier Naranjo-Alcazar, Marta Garcia-Ballesteros, Pedro Zuccarello
https://arxiv.org/abs/2509.14049

Comprehensive Evaluation of CNN-Based Audio Tagging Models on Resource-Constrained Devices
Convolutional Neural Networks (CNNs) have demonstrated exceptional performance in audio tagging tasks. However, deploying these models on resource-constrained devices like the Raspberry Pi poses challenges related to computational efficiency and thermal management. In this paper, a comprehensive evaluation of multiple convolutional neural network (CNN) architectures for audio tagging on the Raspberry Pi is conducted, encompassing all 1D and 2D models from the Pretrained Audio Neural Networks (P…

Accuracy Paradox in Large Language Models: Regulating Hallucination Risks in Generative AI
Zihao Li, Weiwei Yi, Jiahong Chen
https://arxiv.org/abs/2509.13345 https://

Accuracy Paradox in Large Language Models: Regulating Hallucination Risks in Generative AI
As Large Language Models (LLMs) permeate everyday decision-making, their epistemic and societal risks demand urgent scrutiny. Hallucinations, the generation of fabricated, misleading, oversimplified or untrustworthy outputs, has emerged as imperative challenges. While regulatory, academic, and technical discourse position accuracy as the principal benchmark for mitigating such harms, this article contends that overreliance on accuracy misdiagnoses the problem and has counterproductive effect: t…

Ensemble of Pre-Trained Models for Long-Tailed Trajectory Prediction
Divya Thuremella, Yi Yang, Simon Wanna, Lars Kunze, Daniele De Martini
https://arxiv.org/abs/2509.13914 http…

Ensemble of Pre-Trained Models for Long-Tailed Trajectory Prediction
This work explores the application of ensemble modeling to the multidimensional regression problem of trajectory prediction for vehicles in urban environments. As newer and bigger state-of-the-art prediction models for autonomous driving continue to emerge, an important open challenge is the problem of how to combine the strengths of these big models without the need for costly re-training. We show how, perhaps surprisingly, combining state-of-the-art deep learning models out-of-the-box (withou…

PREDICT-GBM: Platform for Robust Evaluation and Development of Individualized Computational Tumor Models in Glioblastoma
L. Zimmer, J. Weidner, M. Balcerak, F. Kofler, I. Ezhov, B. Menze, B. Wiestler
https://arxiv.org/abs/2509.13360

PREDICT-GBM: Platform for Robust Evaluation and Development of Individualized Computational Tumor Models in Glioblastoma
Glioblastoma is the most prevalent primary brain malignancy, distinguished by its highly invasive behavior and exceptionally high rates of recurrence. Conventional radiation therapy, which employs uniform treatment margins, fails to account for patient-specific anatomical and biological factors that critically influence tumor cell migration. To address this limitation, numerous computational models of glioblastoma growth have been developed, enabling generation of tumor cell distribution maps e…

Sample Size Calculations for the Development of Risk Prediction Models that Account for Performance Variability
Menelaos Pavlou, Rumana Z. Omar, Gareth Ambler
https://arxiv.org/abs/2509.14028

Sample Size Calculations for the Development of Risk Prediction Models that Account for Performance Variability
Existing approaches to sample size calculations for developing clinical prediction models have focused on ensuring that the expected value of a chosen performance measure meets a pre-specified target. For example, to limit model-overfitting, the sample size is commonly chosen such that the expected calibration slope (CS) is 0.9, close to 1 for a perfectly calibrated model. In practice, due to sampling variability, model performance can vary considerably across different development samples of t…

HYCO: Hybrid-Cooperative Learning for Data-Driven PDE Modeling
Lorenzo Liverani, Matthys Steynberg, Enrique Zuazua
https://arxiv.org/abs/2509.14123 https://

HYCO: Hybrid-Cooperative Learning for Data-Driven PDE Modeling
We present Hybrid-Cooperative Learning (HYCO), a hybrid modeling framework that iteratively integrates physics-based and data-driven models through a mutual regularization mechanism. Unlike traditional approaches that impose physical constraints directly on synthetic models, HYCO treats the physical and synthetic components as co-trained agents: the physical and synthetic models are nudged toward agreement, while the synthetic model is enhanced to better fit the available data. This cooperative…

Parallelizable Feynman-Kac Models for Universal Probabilistic Programming
Michele Boreale (University of Florence), Luisa Collodi (University of Florence)
https://arxiv.org/abs/2509.14092

Parallelizable Feynman-Kac Models for Universal Probabilistic Programming
We study provably correct and efficient instantiations of Sequential Monte Carlo (SMC) inference in the context of formal operational semantics of Probabilistic Programs (PPs). We focus on universal PPs featuring sampling from arbitrary measures and conditioning/reweighting in unbounded loops. We first equip Probabilistic Program Graphs (PPGs), an automata-theoretic description format of PPs, with an expectation-based semantics over infinite execution traces, which also incorporates trace weigh…

Thermal states emerging from low-entanglement background in disordered spin models
Yule Ma, Qianqian Chen, Mingyang Li, Zlatko Papi\'c, Zheng Zhu
https://arxiv.org/abs/2509.13420

Thermal states emerging from low-entanglement background in disordered spin models
Thermalization in isolated quantum systems is governed by the eigenstate thermalization hypothesis, while strong disorder can induce its breakdown via many-body localization. Here we show that disorder can also generate a narrow band of thermal eigenstates embedded in an otherwise non-thermal spectrum. We illustrate this generic mechanism using paradigmatic spin-1 models, including Heisenberg, XY, and Affleck-Kennedy-Lieb-Tasaki (AKLT) models with several types of disorder. By analyzing their l…

The impact of modeling approaches on controlling safety-critical, highly perturbed systems: the case for data-driven models
Piotr {\L}aszkiewicz, Maria Carvalho, Cl\'audia Soares, Pedro Louren\c{c}o
https://arxiv.org/abs/2509.13531

The impact of modeling approaches on controlling safety-critical, highly perturbed systems: the case for data-driven models
This paper evaluates the impact of three system models on the reference trajectory tracking error of the LQR optimal controller, in the challenging problem of guidance and control of the state of a system under strong perturbations and reconfiguration. We compared a smooth Linear Time Variant system learned from data (DD-LTV) with state of the art Linear Time Variant (LTV) system identification methods, showing its superiority in the task of state propagation. Moreover, we have found that DD-LT…

Evaluating the Limits of QAOA Parameter Transfer at High-Rounds on Sparse Ising Models With Geometrically Local Cubic Terms
Elijah Pelofske, Marek Rams, Andreas B\"artschi, Piotr Czarnik, Paolo Braccia, Lukasz Cincio, Stephan Eidenbenz
https://arxiv.org/abs/2509.13528

Evaluating the Limits of QAOA Parameter Transfer at High-Rounds on Sparse Ising Models With Geometrically Local Cubic Terms
The emergent practical applicability of the Quantum Approximate Optimization Algorithm (QAOA) for approximate combinatorial optimization is a subject of considerable interest. One of the primary limitations of QAOA is the task of finding a set of good parameters. Parameter transfer is a phenomenon where QAOA angles trained on problem instances that are self-similar tend to perform well for other problem instances from that similar class. This suggests a potentially highly efficient and scalable…

@… we need better licensing models across the board

Quelle surprise: seriöse Webseiten blocken den Zugang für KI-Trainingszugriff eher als Seiten, deren Zweck der Desinformation dient.
https://arxiv.org/abs/2510.10315

Cross-diffusion limits in multispecies kinetic models
Ansgar J\"ungel, Annamaria Pollino, Satoshi Taguchi
https://arxiv.org/abs/2509.14046 https://arx…

Cross-diffusion limits in multispecies kinetic models
Cross-diffusion systems are formally derived from multispecies kinetic models in the diffusion limit. The first limit in the multispecies BGK model of Gross and Krook leads to a variant of the non-isothermal Maxwell-Stefan equations. The second limit in a BGK model with Brinkman-type force term yields generalized Busenberg-Travis equations, which reduce for constant temperature to the classical Busenberg-Travis system for segregating population species. Entropy equalities are derived for the ki…

Scrub It Out! Erasing Sensitive Memorization in Code Language Models via Machine Unlearning
Zhaoyang Chu, Yao Wan, Zhikun Zhang, Di Wang, Zhou Yang, Hongyu Zhang, Pan Zhou, Xuanhua Shi, Hai Jin, David Lo
https://arxiv.org/abs/2509.13755

Scrub It Out! Erasing Sensitive Memorization in Code Language Models via Machine Unlearning
While Code Language Models (CLMs) have demonstrated superior performance in software engineering tasks such as code generation and summarization, recent empirical studies reveal a critical privacy vulnerability: these models exhibit unintended memorization of sensitive training data, enabling verbatim reproduction of confidential information when specifically prompted. To address this issue, several approaches, including training data de-duplication and differential privacy augmentation, have b…

A small number of samples can poison LLMs of any size:
https://www.anthropic.com/research/small-samples-poison
"In a joint study with the UK AI Security Institute and the Alan Turing Institute, we found that as few as 250 malicious documents can produce a "

RepCaM : Exploring Transparent Visual Prompt With Inference-Time Re-Parameterization for Neural Video Delivery
Rongyu Zhang, Xize Duan, Jiaming Liu, Li Du, Yuan Du, Dan Wang, Shanghang Zhang, Fangxin Wang
https://arxiv.org/abs/2509.14002

RepCaM++: Exploring Transparent Visual Prompt With Inference-Time Re-Parameterization for Neural Video Delivery
Recently, content-aware methods have been employed to reduce bandwidth and enhance the quality of Internet video delivery. These methods involve training distinct content-aware super-resolution (SR) models for each video chunk on the server, subsequently streaming the low-resolution (LR) video chunks with the SR models to the client. Prior research has incorporated additional partial parameters to customize the models for individual video chunks. However, this leads to parameter accumulation an…

Differential Privacy in Federated Learning: Mitigating Inference Attacks with Randomized Response
Ozer Ozturk, Busra Buyuktanir, Gozde Karatas Baydogmus, Kazim Yildiz
https://arxiv.org/abs/2509.13987

Differential Privacy in Federated Learning: Mitigating Inference Attacks with Randomized Response
Machine learning models used for distributed architectures consisting of servers and clients require large amounts of data to achieve high accuracy. Data obtained from clients are collected on a central server for model training. However, storing data on a central server raises concerns about security and privacy. To address this issue, a federated learning architecture has been proposed. In federated learning, each client trains a local model using its own data. The trained models are periodic…

"People should not "blindly trust" everything AI tools tell them, the boss of Google's parent company Alphabet told the BBC.
In an exclusive interview, chief executive Sundar Pichai said that AI models are "prone to errors" and urged people to use them alongside other tools.
Mr Pichai said it highlighted the importance of having a rich information ecosystem, rather than solely relying on AI technology."

Don't blindly trust what AI tells you, Google boss tells BBC
Sundar Pichai candidly acknowledged concerns about inaccurate answers generated by Google's models.

Always fun/challenging to read new AI (pre)papers like this. "Base models know how to reason, thinking models learn when".
#AI #Google #reasoning

Wardley: Not a Fan of ChatGPT https://blog.bmannconsulting.com/3m3dmlo7t7k2t

Wardley: Not a Fan of ChatGPT - 3lxsr74rnk22h
Large language models are coherence engines not truth engines, we don't teach the critical thinking needed to properly evaluate the outputs.

From Next Token Prediction to (STRIPS) World Models -- Preliminary Results
Carlos N\'u\~nez-Molina, Vicen\c{c} G\'omez, Hector Geffner
https://arxiv.org/abs/2509.13389 h…

From Next Token Prediction to (STRIPS) World Models -- Preliminary Results
We consider the problem of learning propositional STRIPS world models from action traces alone, using a deep learning architecture (transformers) and gradient descent. The task is cast as a supervised next token prediction problem where the tokens are the actions, and an action $a$ may follow an action sequence if the hidden effects of the previous actions do not make an action precondition of $a$ false. We show that a suitable transformer architecture can faithfully represent propositional STR…

Designing Psychometric Bias Measures for ChatBots: An Application to Racial Bias Measurement
Mouhacine Benosman
https://arxiv.org/abs/2509.13324 https://ar…

Designing Psychometric Bias Measures for ChatBots: An Application to Racial Bias Measurement
Artificial intelligence (AI), particularly in the form of large language models (LLMs) or chatbots, has become increasingly integrated into our daily lives. In the past five years, several LLMs have been introduced, including ChatGPT by OpenAI, Claude by Anthropic, and Llama by Meta, among others. These models have the potential to be employed across a wide range of human-machine interaction applications, such as chatbots for information retrieval, assistance in corporate hiring decisions, coll…

"By examining several widely used AI models designed for cancer detection, the researchers found that performance varied depending on patients’ self-reported gender, race, and age. They also identified multiple reasons why these disparities occur."
What AI Learned From Cancer Slides Shocked Researchers
https://scitechdaily.com/what-ai-learned-from-cancer-slides-shocked-researchers/

Diving into Mitigating Hallucinations from a Vision Perspective for Large Vision-Language Models
Weihang Wang, Xinhao Li, Ziyue Wang, Yan Pang, Jielei Zhang, Peiyi Li, Qiang Zhang, Longwen Gao
https://arxiv.org/abs/2509.13836

Diving into Mitigating Hallucinations from a Vision Perspective for Large Vision-Language Models
Object hallucination in Large Vision-Language Models (LVLMs) significantly impedes their real-world applicability. As the primary component for accurately interpreting visual information, the choice of visual encoder is pivotal. We hypothesize that the diverse training paradigms employed by different visual encoders instill them with distinct inductive biases, which leads to their diverse hallucination performances. Existing benchmarks typically focus on coarse-grained hallucination detection a…

Do Large Language Models Understand Word Senses?
Domenico Meconi, Simone Stirpe, Federico Martelli, Leonardo Lavalle, Roberto Navigli
https://arxiv.org/abs/2509.13905 https://…

Do Large Language Models Understand Word Senses?
Understanding the meaning of words in context is a fundamental capability for Large Language Models (LLMs). Despite extensive evaluation efforts, the extent to which LLMs show evidence that they truly grasp word senses remains underexplored. In this paper, we address this gap by evaluating both i) the Word Sense Disambiguation (WSD) capabilities of instruction-tuned LLMs, comparing their performance to state-of-the-art systems specifically designed for the task, and ii) the ability of two top-p…

TICL: Text-Embedding KNN For Speech In-Context Learning Unlocks Speech Recognition Abilities of Large Multimodal Models
Haolong Zheng, Yekaterina Yegorova, Mark Hasegawa-Johnson
https://arxiv.org/abs/2509.13395

TICL: Text-Embedding KNN For Speech In-Context Learning Unlocks Speech Recognition Abilities of Large Multimodal Models
Speech foundation models have recently demonstrated the ability to perform Speech In-Context Learning (SICL). Selecting effective in-context examples is crucial for SICL performance, yet selection methodologies remain underexplored. In this work, we propose Text-Embedding KNN for SICL (TICL), a simple pipeline that uses semantic context to enhance off-the-shelf large multimodal models' speech recognition ability without fine-tuning. Across challenging automatic speech recognition tasks, includi…

New OpenAI models likely to pose "high" cybersecurity risk
OpenAI says the cyber capabilities of its frontier AI models are accelerating and warns Wednesday that upcoming models are likely to pose a "high" risk […]
😶‍🌫️ https://www.axios.com/2025/12/10/openai-ne

Exclusive: New OpenAI models likely pose "high" cybersecurity risk, company says
OpenAI says it's stepping up efforts to protect against attacks.

The threat of analytic flexibility in using large language models to simulate human data: A call to attention
Jamie Cummins
https://arxiv.org/abs/2509.13397 https://

The threat of analytic flexibility in using large language models to simulate human data: A call to attention
Social scientists are now using large language models to create "silicon samples" - synthetic datasets intended to stand in for human respondents, aimed at revolutionising human subjects research. However, there are many analytic choices which must be made to produce these samples. Though many of these choices are defensible, their impact on sample quality is poorly understood. I map out these analytic choices and demonstrate how a very small number of decisions can dramatically change the corr…

A look at Apple's iPhone roadmap; sources say the iPhone 18 Pro lineup will get under-screen Face ID, and the front camera will move to the top-left corner (The Information)
https://www.theinformation.com/articles/inside-apples-ip…

Inside Apple’s iPhone Road Map, From Foldable Screens to Curved Glass Cases
The iPhone family is growing again. Over the next two years, Apple is planning a major expansion of its smartphone product line that could see it selling at least seven new models of its flagship device by fall 2027, up from five iPhone models today, according to multiple people working directly ...

Are Prompts All You Need? Evaluating Prompt-Based Large Language Models (LLM)s for Software Requirements Classification
Manal Binkhonain, Reem Alfayaz
https://arxiv.org/abs/2509.13868

Are Prompts All You Need? Evaluating Prompt-Based Large Language Models (LLM)s for Software Requirements Classification
Requirements classification assigns natural language requirements to predefined classes, such as functional and non functional. Accurate classification reduces risk and improves software quality. Most existing models rely on supervised learning, which needs large labeled data that are costly, slow to create, and domain dependent; they also generalize poorly and often require retraining for each task. This study tests whether prompt based large language models can reduce data needs. We benchmark…

Language models' activations linearly encode training-order recency
Dmitrii Krasheninnikov, Richard E. Turner, David Krueger
https://arxiv.org/abs/2509.14223 https://…

Language models' activations linearly encode training-order recency
We show that language models' activations linearly encode when information was learned during training. Our setup involves creating a model with a known training order by sequentially fine-tuning Llama-3.2-1B on six disjoint but otherwise similar datasets about named entities. We find that the average activations of test samples for the six training datasets encode the training order: when projected into a 2D subspace, these centroids are arranged exactly in the order of training and lie on a s…

Early Stopping Chain-of-thoughts in Large Language Models
Minjia Mao, Bowen Yin, Yu Zhu, Xiao Fang
https://arxiv.org/abs/2509.14004 https://arxiv.org/pdf/2…

Early Stopping Chain-of-thoughts in Large Language Models
Reasoning large language models (LLMs) have demonstrated superior capacities in solving complicated problems by generating long chain-of-thoughts (CoT), but such a lengthy CoT incurs high inference costs. In this study, we introduce ES-CoT, an inference-time method that shortens CoT generation by detecting answer convergence and stopping early with minimal performance loss. At the end of each reasoning step, we prompt the LLM to output its current final answer, denoted as a step answer. We then…

OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft
Zihao Wang, Muyao Li, Kaichen He, Xiangyu Wang, Zhancun Mu, Anji Liu, Yitao Liang
https://arxiv.org/abs/2509.13347

OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft
The choice of action spaces is a critical yet unresolved challenge in developing capable, end-to-end trainable agents. This paper first presents a large-scale, systematic comparison of prominent abstracted action spaces and tokenizers for Vision-Language-Action (VLA) or hierarchical agent models in the open-ended Minecraft. Our analysis reveals that no single action space is universally optimal; instead, the most effective abstraction is highly task-dependent, creating a dilemma for building ge…

A tree-based Polynomial Chaos expansion for surrogate modeling and sensitivity analysis of complex numerical models
Faten Ben Said (CERMICS, EDF R\&D LNHE), Aur\'elien Alfonsi (CERMICS, MATHRISK), Anne Dutfoy (EDF R\&D PERICLES), C\'edric Goeury (EDF R\&D LNHE, LHSV), Magali Jodeau (EDF R\&D LNHE, LHSV), Julien Reygner (CERMICS, RT-UQ), Fabrice Zaoui (EDF R\&D LNHE)

A tree-based Polynomial Chaos expansion for surrogate modeling and sensitivity analysis of complex numerical models
This paper introduces Tree-based Polynomial Chaos Expansion (Tree-PCE), a novel surrogate modeling technique designed to efficiently approximate complex numerical models exhibiting nonlinearities and discontinuities. Tree-PCE combines the expressive power of Polynomial Chaos Expansion (PCE) with an adaptive partitioning strategy inspired by regression trees. By recursively dividing the input space into hyperrectangular subdomains and fitting localized PCEs, Tree-PCE constructs a piecewise polyn…

PhysicalAgent: Towards General Cognitive Robotics with Foundation World Models
Artem Lykov, Jeffrin Sam, Hung Khang Nguyen, Vladislav Kozlovskiy, Yara Mahmoud, Valerii Serpiva, Miguel Altamirano Cabrera, Mikhail Konenkov, Dzmitry Tsetserukou
https://arxiv.org/abs/2509.13903

PhysicalAgent: Towards General Cognitive Robotics with Foundation World Models
We introduce PhysicalAgent, an agentic framework for robotic manipulation that integrates iterative reasoning, diffusion-based video generation, and closed-loop execution. Given a textual instruction, our method generates short video demonstrations of candidate trajectories, executes them on the robot, and iteratively re-plans in response to failures. This approach enables robust recovery from execution errors. We evaluate PhysicalAgent across multiple perceptual modalities (egocentric, third-p…

End-to-End Complexity Analysis for Quantum Simulation of the Extended Jaynes-Cummings Models
Nam Nguyen, Michael Yu, Alan Robertson, Hiromichi Nishimura, Samuel J. Elman, Benjamin Koltenbah
https://arxiv.org/abs/2509.13546

End-to-End Complexity Analysis for Quantum Simulation of the Extended Jaynes-Cummings Models
The extended Jaynes-Cummings model (eJCM) is a foundational framework for describing multi-mode light-matter interactions, with direct applications in quantum technologies such as photon addition and quasi-noiseless amplification. However, the model's complexity makes classical simulation intractable for large systems that could be of practical interest. In this work, we present a comprehensive, end-to-end framework for the quantum simulation of the eJCM. We develop explicit quantum algorithms …

Masked Diffusion Models as Energy Minimization
Sitong Chen, Shen Nie, Jiacheng Sun, Zijin Feng, Zhenguo Li, Ji-Rong Wen, Chongxuan Li
https://arxiv.org/abs/2509.13866 https://…

Masked Diffusion Models as Energy Minimization
We present a systematic theoretical framework that interprets masked diffusion models (MDMs) as solutions to energy minimization problems in discrete optimal transport. Specifically, we prove that three distinct energy formulations--kinetic, conditional kinetic, and geodesic energy--are mathematically equivalent under the structure of MDMs, and that MDMs minimize all three when the mask schedule satisfies a closed-form optimality condition. This unification not only clarifies the theoretical fo…

Integrating Text and Time-Series into (Large) Language Models to Predict Medical Outcomes
Iyadh Ben Cheikh Larbi, Ajay Madhavan Ravichandran, Aljoscha Burchardt, Roland Roller
https://arxiv.org/abs/2509.13696

Integrating Text and Time-Series into (Large) Language Models to Predict Medical Outcomes
Large language models (LLMs) excel at text generation, but their ability to handle clinical classification tasks involving structured data, such as time series, remains underexplored. In this work, we adapt instruction-tuned LLMs using DSPy-based prompt optimization to process clinical notes and structured EHR inputs jointly. Our results show that this approach achieves performance on par with specialized multimodal systems while requiring less complexity and offering greater adaptability acros…

Consistent View Alignment Improves Foundation Models for 3D Medical Image Segmentation
Puru Vaish, Felix Meister, Tobias Heimann, Christoph Brune, Jelmer M. Wolterink
https://arxiv.org/abs/2509.13846

Consistent View Alignment Improves Foundation Models for 3D Medical Image Segmentation
Many recent approaches in representation learning implicitly assume that uncorrelated views of a data point are sufficient to learn meaningful representations for various downstream tasks. In this work, we challenge this assumption and demonstrate that meaningful structure in the latent space does not emerge naturally. Instead, it must be explicitly induced. We propose a method that aligns representations from different views of the data to align complementary information without inducing false…

Crash Report Enhancement with Large Language Models: An Empirical Study
S M Farah Al Fahim (Peter), Md Nakhla Rafi (Peter), Zeyang Ma (Peter), Dong Jae Kim (Peter), Tse-Hsun (Peter), Chen
https://arxiv.org/abs/2509.13535

Crash Report Enhancement with Large Language Models: An Empirical Study
Crash reports are central to software maintenance, yet many lack the diagnostic detail developers need to debug efficiently. We examine whether large language models can enhance crash reports by adding fault locations, root-cause explanations, and repair suggestions. We study two enhancement strategies: Direct-LLM, a single-shot approach that uses stack-trace context, and Agentic-LLM, an iterative approach that explores the repository for additional evidence. On a dataset of 492 real-world cras…

Graph-Regularized Learning of Gaussian Mixture Models
Shamsiiat Abdurakhmanova, Alex Jung
https://arxiv.org/abs/2509.13855 https://arxiv.org/pdf/2509.13855…

Graph-Regularized Learning of Gaussian Mixture Models
We present a graph-regularized learning of Gaussian Mixture Models (GMMs) in distributed settings with heterogeneous and limited local data. The method exploits a provided similarity graph to guide parameter sharing among nodes, avoiding the transfer of raw data. The resulting model allows for flexible aggregation of neighbors' parameters and outperforms both centralized and locally trained GMMs in heterogeneous, low-sample regimes.

FlightDiffusion: Revolutionising Autonomous Drone Training with Diffusion Models Generating FPV Video
Valerii Serpiva, Artem Lykov, Faryal Batool, Vladislav Kozlovskiy, Miguel Altamirano Cabrera, Dzmitry Tsetserukou
https://arxiv.org/abs/2509.14082

FlightDiffusion: Revolutionising Autonomous Drone Training with Diffusion Models Generating FPV Video
We present FlightDiffusion, a diffusion-model-based framework for training autonomous drones from first-person view (FPV) video. Our model generates realistic video sequences from a single frame, enriched with corresponding action spaces to enable reasoning-driven navigation in dynamic environments. Beyond direct policy learning, FlightDiffusion leverages its generative capabilities to synthesize diverse FPV trajectories and state-action pairs, facilitating the creation of large-scale training …

Large Language Models Discriminate Against Speakers of German Dialects
Minh Duc Bui, Carolin Holtermann, Valentin Hofmann, Anne Lauscher, Katharina von der Wense
https://arxiv.org/abs/2509.13835

Large Language Models Discriminate Against Speakers of German Dialects
Dialects represent a significant component of human culture and are found across all regions of the world. In Germany, more than 40% of the population speaks a regional dialect (Adler and Hansen, 2022). However, despite cultural importance, individuals speaking dialects often face negative societal stereotypes. We examine whether such stereotypes are mirrored by large language models (LLMs). We draw on the sociolinguistic literature on dialect perception to analyze traits commonly associated wi…

Interleaving Natural Language Prompting with Code Editing for Solving Programming Tasks with Generative AI Models
Victor-Alexandru P\u{a}durean, Paul Denny, Andrew Luxton-Reilly, Alkis Gotovos, Adish Singla
https://arxiv.org/abs/2509.14088

Interleaving Natural Language Prompting with Code Editing for Solving Programming Tasks with Generative AI Models
Nowadays, computing students often rely on both natural-language prompting and manual code editing to solve programming tasks. Yet we still lack a clear understanding of how these two modes are combined in practice, and how their usage varies with task complexity and student ability. In this paper, we investigate this through a large-scale study in an introductory programming course, collecting 13,305 interactions from 355 students during a three-day laboratory activity. Our analysis shows that…

NIRVANA: Structured pruning reimagined for large language models compression
Mengting Ai, Tianxin Wei, Sirui Chen, Jingrui He
https://arxiv.org/abs/2509.14230 https://

NIRVANA: Structured pruning reimagined for large language models compression
Structured pruning of large language models (LLMs) offers substantial efficiency improvements by removing entire hidden units, yet current approaches often suffer from significant performance degradation, particularly in zero-shot settings, and necessitate costly recovery techniques such as supervised fine-tuning (SFT) or adapter insertion. To address these critical shortcomings, we introduce NIRVANA, a novel pruning method explicitly designed to balance immediate zero-shot accuracy preservatio…

Inspections by Taiwan's National Security Bureau (NSB) of five Chinese generative AI apps -- Deepseek, Doubao (豆包), Yiyan (文心一言), Tongyi (通義千問), and Yuanbao (騰訊元寶) -- found violations of users' communication security across several indicators.
https://focustaiwan.tw/cross-strait/2025…

Taiwan warns of biases, data breach in Deepseek, other Chinese AI - Focus Taiwan
The National Security Bureau (NSB) has urged Taiwanese people to be on alert when using Chinese generative artificial intelligence (AI) language models due to potential security breaches and the spread of "disinformation" following recent inspections of five such apps.

Evolution of meta's llama models and parameter-efficient fine-tuning of large language models: a survey
Abdulhady Abas Abdullah, Arkaitz Zubiaga, Seyedali Mirjalili, Amir H. Gandomi, Fatemeh Daneshfar, Mohammadsadra Amini, Alan Salam Mohammed, Hadi Veisi
https://arxiv.org/abs/2510.12178

Evolution of meta's llama models and parameter-efficient fine-tuning of large language models: a survey
This review surveys the rapid evolution of Meta AI's LLaMA (Large Language Model Meta AI) series - from LLaMA 1 through LLaMA 4 and the specialized parameter-efficient fine-tuning (PEFT) methods developed for these models. We first describe the LLaMA family of foundation models (7B-65B to 288B parameters), their architectures (including native multimodal and Mixtureof-Experts variants), and key performance characteristics. We then describe and discuss the concept of PEFT, which adapts large pre…

Hala Technical Report: Building Arabic-Centric Instruction & Translation Models at Scale
Hasan Abed Al Kader Hammoud, Mohammad Zbeeb, Bernard Ghanem
https://arxiv.org/abs/2509.14008

Hala Technical Report: Building Arabic-Centric Instruction & Translation Models at Scale
We present Hala, a family of Arabic-centric instruction and translation models built with our translate-and-tune pipeline. We first compress a strong AR$\leftrightarrow$EN teacher to FP8 (yielding $\sim$2$\times$ higher throughput with no quality loss) and use it to create high-fidelity bilingual supervision. A lightweight language model LFM2-1.2B is then fine-tuned on this data and used to translate high-quality English instruction sets into Arabic, producing a million-scale corpus tailored to…

OpenAI launches FrontierScience, a benchmark to measure models' expert-level scientific reasoning with 700 questions, finding GPT-5.2 is its strongest model (OpenAI)
https://openai.com/index/frontierscience/

ST-LINK: Spatially-Aware Large Language Models for Spatio-Temporal Forecasting
Hyotaek Jeon, Hyunwook Lee, Juwon Kim, Sungahn Ko
https://arxiv.org/abs/2509.13753 https://…

ST-LINK: Spatially-Aware Large Language Models for Spatio-Temporal Forecasting
Traffic forecasting represents a crucial problem within intelligent transportation systems. In recent research, Large Language Models (LLMs) have emerged as a promising method, but their intrinsic design, tailored primarily for sequential token processing, introduces notable challenges in effectively capturing spatial dependencies. Specifically, the inherent limitations of LLMs in modeling spatial relationships and their architectural incompatibility with graph-structured spatial data remain la…

Pre-Manipulation Alignment Prediction with Parallel Deep State-Space and Transformer Models
Motonari Kambara, Komei Sugiura
https://arxiv.org/abs/2509.13839 https://

Pre-Manipulation Alignment Prediction with Parallel Deep State-Space and Transformer Models
In this work, we address the problem of predicting the future success of open-vocabulary object manipulation tasks. Conventional approaches typically determine success or failure after the action has been carried out. However, they make it difficult to prevent potential hazards and rely on failures to trigger replanning, thereby reducing the efficiency of object manipulation sequences. To overcome these challenges, we propose a model, which predicts the alignment between a pre-manipulation egoc…

Benchmarking foundation models for hyperspectral image classification: Application to cereal crop type mapping
Walid Elbarz, Mohamed Bourriz, Hicham Hajji, Hamd Ait Abdelali, Fran\c{c}ois Bourzeix
https://arxiv.org/abs/2510.11576

Benchmarking foundation models for hyperspectral image classification: Application to cereal crop type mapping
Foundation models are transforming Earth observation, but their potential for hyperspectral crop mapping remains underexplored. This study benchmarks three foundation models for cereal crop mapping using hyperspectral imagery: HyperSigma, DOFA, and Vision Transformers pre-trained on the SpectralEarth dataset (a large multitemporal hyperspectral archive). Models were fine-tuned on manually labeled data from a training region and evaluated on an independent test region. Performance was measured w…

Automated Triaging and Transfer Learning of Incident Learning Safety Reports Using Large Language Representational Models
Peter Beidler, Mark Nguyen, Kevin Lybarger, Ola Holmberg, Eric Ford, John Kang
https://arxiv.org/abs/2509.13706

Automated Triaging and Transfer Learning of Incident Learning Safety Reports Using Large Language Representational Models
PURPOSE: Incident reports are an important tool for safety and quality improvement in healthcare, but manual review is time-consuming and requires subject matter expertise. Here we present a natural language processing (NLP) screening tool to detect high-severity incident reports in radiation oncology across two institutions. METHODS AND MATERIALS: We used two text datasets to train and evaluate our NLP models: 7,094 reports from our institution (Inst.), and 571 from IAEA SAFRON (SF), all of …

TFMAdapter: Lightweight Instance-Level Adaptation of Foundation Models for Forecasting with Covariates
Afrin Dange, Sunita Sarawagi
https://arxiv.org/abs/2509.13906 https://

TFMAdapter: Lightweight Instance-Level Adaptation of Foundation Models for Forecasting with Covariates
Time Series Foundation Models (TSFMs) have recently achieved state-of-the-art performance in univariate forecasting on new time series simply by conditioned on a brief history of past values. Their success demonstrates that large-scale pretraining across diverse domains can acquire the inductive bias to generalize from temporal patterns in a brief history. However, most TSFMs are unable to leverage covariates -- future-available exogenous variables critical for accurate forecasting in many appl…

Source: OpenAI rolled back ChatGPT's model router, which sent some queries to reasoning models, for Free and $5/month Go tiers, as it was costly and hurt DAUs (Maxwell Zeff/Wired)
https://www.wired.com/story/openai-router-relaunch-gpt-5-sam-altman/

OpenAI Rolls Back ChatGPT’s Model Router System for Most Users
As OpenAI scrambles to improve ChatGPT, it's ditching a feature in its free tier that contributed to last summer's user revolt.

Explicit Reasoning Makes Better Judges: A Systematic Study on Accuracy, Efficiency, and Robustness
Pratik Jayarao, Himanshu Gupta, Neeraj Varshney, Chaitanya Dwivedi
https://arxiv.org/abs/2509.13332

Explicit Reasoning Makes Better Judges: A Systematic Study on Accuracy, Efficiency, and Robustness
As Large Language Models (LLMs) are increasingly adopted as automated judges in benchmarking and reward modeling, ensuring their reliability, efficiency, and robustness has become critical. In this work, we present a systematic comparison of "thinking" and "non-thinking" LLMs in the LLM-as-a-judge paradigm using open-source Qwen 3 models of relatively small sizes (0.6B, 1.7B, and 4B parameters). We evaluate both accuracy and computational efficiency (FLOPs) on RewardBench tasks, and further exa…

Evaluating Classical Software Process Models as Coordination Mechanisms for LLM-Based Software Generation
Duc Minh Ha, Phu Trac Kien, Tho Quan, Anh Nguyen-Duc
https://arxiv.org/abs/2509.13942

Evaluating Classical Software Process Models as Coordination Mechanisms for LLM-Based Software Generation
[Background] Large Language Model (LLM)-based multi-agent systems (MAS) are transforming software development by enabling autonomous collaboration. Classical software processes such asWaterfall, V-Model, and Agile offer structured coordination patterns that can be repurposed to guide these agent interactions. [Aims] This study explores how traditional software development processes can be adapted as coordination scaffolds for LLM based MAS and examines their impact on code quality, cost, and pr…

Crosslisted article(s) found for cs.CY. https://arxiv.org/list/cs.CY/new
[1/1]:
- All Models Are Wrong, But Can They Be Useful? Lessons from COVID-19 Agent-Based Models: A Systema...
Von Hoene, Von Hoene, Peter, Hopson, Csizmadia, Fenyk, Barner, Leslie, Kavak, Zufle, Roess, Anderson
…

Multi-robot Multi-source Localization in Complex Flows with Physics-Preserving Environment Models
Benjamin Shaffer, Victoria Edwards, Brooks Kinch, Nathaniel Trask, M. Ani Hsieh
https://arxiv.org/abs/2509.14228

Multi-robot Multi-source Localization in Complex Flows with Physics-Preserving Environment Models
Source localization in a complex flow poses a significant challenge for multi-robot teams tasked with localizing the source of chemical leaks or tracking the dispersion of an oil spill. The flow dynamics can be time-varying and chaotic, resulting in sporadic and intermittent sensor readings, and complex environmental geometries further complicate a team's ability to model and predict the dispersion. To accurately account for the physical processes that drive the dispersion dynamics, robots must…

State Space Models over Directed Graphs
Junzhi She, Xunkai Li, Rong-Hua Li, Guoren Wang
https://arxiv.org/abs/2509.13735 https://arxiv.org/pdf/2509.13735…

State Space Models over Directed Graphs
Directed graphs are ubiquitous across numerous domains, where the directionality of edges encodes critical causal dependencies. However, existing GNNs and graph Transformers tailored for directed graphs face two major challenges: (1) effectively capturing long-range causal dependencies derived from directed edges; (2) balancing accuracy and training efficiency when processing large-scale graph datasets. In recent years, state space models (SSMs) have achieved substantial progress in causal sequ…

Apertus: Democratizing Open and Compliant LLMs for Global Language Environments
Alejandro Hern\'andez-Cano, Alexander H\"agele, Allen Hao Huang, Angelika Romanou, Antoni-Joan Solergibert, Barna Pasztor, Bettina Messmer, Dhia Garbaya, Eduard Frank \v{D}urech, Ido Hakimi, Juan Garc\'ia Giraldo, Mete Ismayilzada, Negar Foroutan, Skander Moalla, Tiancheng Chen, Vinko Sabol\v{c}ec, Yixuan Xu, Michael Aerni, Badr AlKhamissi, Ines Altemir Marinas, Mohammad Hossein Amani, Matin An…

Apertus: Democratizing Open and Compliant LLMs for Global Language Environments
We present Apertus, a fully open suite of large language models (LLMs) designed to address two systemic shortcomings in today's open model ecosystem: data compliance and multilingual representation. Unlike many prior models that release weights without reproducible data pipelines or regard for content-owner rights, Apertus models are pretrained exclusively on openly available data, retroactively respecting robots.txt exclusions and filtering for non-permissive, toxic, and personally identifiabl…

CrowdAgent: Multi-Agent Managed Multi-Source Annotation System
Maosheng Qin, Renyu Zhu, Mingxuan Xia, Chenkai Chen, Zhen Zhu, Minmin Lin, Junbo Zhao, Lu Xu, Changjie Fan, Runze Wu, Haobo Wang
https://arxiv.org/abs/2509.14030

CrowdAgent: Multi-Agent Managed Multi-Source Annotation System
High-quality annotated data is a cornerstone of modern Natural Language Processing (NLP). While recent methods begin to leverage diverse annotation sources-including Large Language Models (LLMs), Small Language Models (SLMs), and human experts-they often focus narrowly on the labeling step itself. A critical gap remains in the holistic process control required to manage these sources dynamically, addressing complex scheduling and quality-cost trade-offs in a unified manner. Inspired by real-wor…

The Allen Institute of AI launches Bolmo 7B and Bolmo 1B, claiming they are "the first fully open byte-level language models", built on its Olmo 3 models (Emilia David/VentureBeat)
https://venturebeat.com/ai/bolmos-architecture-unlock…

From Distributional to Quantile Neural Basis Models: the case of Electricity Price Forecasting
Alessandro Brusaferri, Danial Ramin, Andrea Ballarino
https://arxiv.org/abs/2509.14113

From Distributional to Quantile Neural Basis Models: the case of Electricity Price Forecasting
While neural networks are achieving high predictive accuracy in multi-horizon probabilistic forecasting, understanding the underlying mechanisms that lead to feature-conditioned outputs remains a significant challenge for forecasters. In this work, we take a further step toward addressing this critical issue by introducing the Quantile Neural Basis Model, which incorporates the interpretability principles of Quantile Generalized Additive Models into an end-to-end neural network training framewo…

DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving
Yingyan Li, Shuyao Shang, Weisong Liu, Bing Zhan, Haochen Wang, Yuqi Wang, Yuntao Chen, Xiaoman Wang, Yasong An, Chufeng Tang, Lu Hou, Lue Fan, Zhaoxiang Zhang
https://arxiv.org/abs/2510.12796

DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving
Scaling Vision-Language-Action (VLA) models on large-scale data offers a promising path to achieving a more generalized driving intelligence. However, VLA models are limited by a ``supervision deficit'': the vast model capacity is supervised by sparse, low-dimensional actions, leaving much of their representational power underutilized. To remedy this, we propose \textbf{DriveVLA-W0}, a training paradigm that employs world modeling to predict future images. This task generates a dense, self-supe…

TGPO: Tree-Guided Preference Optimization for Robust Web Agent Reinforcement Learning
Ziyuan Chen, Zhenghui Zhao, Zhangye Han, Miancan Liu, Xianhang Ye, Yiqing Li, Hongbo Min, Jinkui Ren, Xiantao Zhang, Guitao Cao
https://arxiv.org/abs/2509.14172

TGPO: Tree-Guided Preference Optimization for Robust Web Agent Reinforcement Learning
With the rapid advancement of large language models and vision-language models, employing large models as Web Agents has become essential for automated web interaction. However, training Web Agents with reinforcement learning faces critical challenges including credit assignment misallocation, prohibitively high annotation costs, and reward sparsity. To address these issues, we propose Tree-Guided Preference Optimization (TGPO), an offline reinforcement learning framework that proposes a tree-s…

Teaching LLMs to Plan: Logical Chain-of-Thought Instruction Tuning for Symbolic Planning
Pulkit Verma, Ngoc La, Anthony Favier, Swaroop Mishra, Julie A. Shah
https://arxiv.org/abs/2509.13351

Teaching LLMs to Plan: Logical Chain-of-Thought Instruction Tuning for Symbolic Planning
Large language models (LLMs) have demonstrated impressive capabilities across diverse tasks, yet their ability to perform structured symbolic planning remains limited, particularly in domains requiring formal representations like the Planning Domain Definition Language (PDDL). In this paper, we present a novel instruction tuning framework, PDDL-Instruct, designed to enhance LLMs' symbolic planning capabilities through logical chain-of-thought reasoning. Our approach focuses on teaching models t…

Mind the Ethics! The Overlooked Ethical Dimensions of GenAI in Software Modeling Education
Shalini Chakraborty, Lola Burgue\~no, Nathalie Moreno, Javier Troya, Paula Mu\~noz
https://arxiv.org/abs/2509.13896

Mind the Ethics! The Overlooked Ethical Dimensions of GenAI in Software Modeling Education
Generative Artificial Intelligence (GenAI) is rapidly gaining momentum in software modeling education, embraced by both students and educators. As GenAI assists with interpreting requirements, formalizing models, and translating students' mental models into structured notations, it increasingly shapes core learning outcomes such as domain comprehension, diagrammatic thinking, and modeling fluency without clear ethical oversight or pedagogical guidelines. Yet, the ethical implications of this in…

Sources: Amazon is in talks to invest $10B in OpenAI at a $500B valuation, with OpenAI using AWS Trainium chips; Microsoft keeps rights to sell OpenAI models (The Information)
https://www.theinformation.com/articles/openai-talks-raise-least-1…

OpenAI in Talks to Raise At Least $10 Billion From Amazon and Use Its AI Chips
Amazon is in talks to invest $10 billion or more in OpenAI, according to three people familiar with the discussions. The valuation would be higher than $500 billion, one of the people said. The Amazon investment would help OpenAI afford some of the commitments it has made to rent servers from ...

Imagined Autocurricula
Ahmet H. G\"uzel, Matthew Thomas Jackson, Jarek Luca Liesen, Tim Rockt\"aschel, Jakob Nicolaus Foerster, Ilija Bogunovic, Jack Parker-Holder
https://arxiv.org/abs/2509.13341

Imagined Autocurricula
Training agents to act in embodied environments typically requires vast training data or access to accurate simulation, neither of which exists for many cases in the real world. Instead, world models are emerging as an alternative leveraging offline, passively collected data, they make it possible to generate diverse worlds for training agents in simulation. In this work, we harness world models to generate imagined environments to train robust agents capable of generalizing to novel task varia…

Personalized Federated Fine-Tuning of Vision Foundation Models for Healthcare
Adam Tupper, Christian Gagn\'e
https://arxiv.org/abs/2510.12741 https://a…

Personalized Federated Fine-Tuning of Vision Foundation Models for Healthcare
Foundation models open up new possibilities for the use of AI in healthcare. However, even when pre-trained on health data, they still need to be fine-tuned for specific downstream tasks. Furthermore, although foundation models reduce the amount of training data required to achieve good performance, obtaining sufficient data is still a challenge. This is due, in part, to restrictions on sharing and aggregating data from different sources to protect patients' privacy. One possible solution to th…

Google releases Cell2Sentence-Scale 27B (C2S-Scale), a 27B-parameter foundation model for single-cell analysis built on its Gemma family of open models (The Keyword)
https://blog.google/technology/ai/google-gemma-ai-cancer-therapy-discovery/

How a Gemma model helped discover a new potential cancer therapy pathway
We’re launching a new 27 billion parameter foundation model for single-cell analysis built on the Gemma family of open models.

Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST
Monica Sekoyan, Nithin Rao Koluguri, Nune Tadevosyan, Piotr Zelasko, Travis Bartley, Nick Karpov, Jagadeesh Balam, Boris Ginsburg
https://arxiv.org/abs/2509.14128

Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and AST
This report introduces Canary-1B-v2, a fast, robust multilingual model for Automatic Speech Recognition (ASR) and Speech-to-Text Translation (AST). Built with a FastConformer encoder and Transformer decoder, it supports 25 languages primarily European. The model was trained on 1.7M hours of total data samples, including Granary and NeMo ASR Set 3.0, with non-speech audio added to reduce hallucinations for ASR and AST. We describe its two-stage pre-training and fine-tuning process with dynamic d…

Evaluating End-User Device Energy Models in Sustainability Reporting of Browser-Based Web Services
Maja H. Kirkeby, Timmie Lagermann
https://arxiv.org/abs/2510.12566 https://

Evaluating End-User Device Energy Models in Sustainability Reporting of Browser-Based Web Services
Sustainability reporting in web-based services increasingly relies on simplified energy and carbon models such as the Danish Agency of Digital Government's Digst framework and the United Kingdom-based DIMPACT model. Although these models are widely adopted, their accuracy and precision remain underexplored. This paper presents an empirical study evaluating how well such models reflect actual energy consumption during realistic user interactions with common website categories. Energy use was mea…

Anthropic's Skills for Claude, which are conceptually very simple, may become a bigger deal than MCP, whose high token usage is its most significant limitation (Simon Willison/Simon Willison's Weblog)
https://simonwillison.net/2025/Oct/16/claude-skills/

Claude Skills are awesome, maybe a bigger deal than MCP
Anthropic this morning introduced Claude Skills, a new pattern for making new abilities available to their models: Claude can now use Skills to improve how it performs specific tasks. Skills …

DSCC-HS: A Dynamic Self-Reinforcing Framework for Hallucination Suppression in Large Language Models
Xiao Zheng
https://arxiv.org/abs/2509.13702 https://ar…

DSCC-HS: A Dynamic Self-Reinforcing Framework for Hallucination Suppression in Large Language Models
Large Language Model (LLM) hallucination is a significant barrier to their reliable deployment. Current methods like Retrieval-Augmented Generation (RAG) are often reactive. We introduce **Dynamic Self-reinforcing Calibration for Hallucination Suppression (DSCC-HS)**, a novel, proactive framework that intervenes during autoregressive decoding. Inspired by dual-process cognitive theory, DSCC-HS uses a compact proxy model, trained in adversarial roles as a Factual Alignment Proxy (FAP) and a Hall…

On the Use of Hierarchical Vision Foundation Models for Low-Cost Human Mesh Recovery and Pose Estimation
Shuhei Tarashima, Yushan Wang, Norio Tagawa
https://arxiv.org/abs/2510.12660

On the Use of Hierarchical Vision Foundation Models for Low-Cost Human Mesh Recovery and Pose Estimation
In this work, we aim to develop simple and efficient models for human mesh recovery (HMR) and its predecessor task, human pose estimation (HPE). State-of-the-art HMR methods, such as HMR2.0 and its successors, rely on large, non-hierarchical vision transformers as encoders, which are inherited from the corresponding HPE models like ViTPose. To establish baselines across varying computational budgets, we first construct three lightweight HMR2.0 variants by adapting the corresponding ViTPose mode…

You Are What You Train: Effects of Data Composition on Training Context-aware Machine Translation Models
Pawe{\l} M\k{a}ka, Yusuf Can Semerci, Jan Scholtes, Gerasimos Spanakis
https://arxiv.org/abs/2509.14031

You Are What You Train: Effects of Data Composition on Training Context-aware Machine Translation Models
Achieving human-level translations requires leveraging context to ensure coherence and handle complex phenomena like pronoun disambiguation. Sparsity of contextually rich examples in the standard training data has been hypothesized as the reason for the difficulty of context utilization. In this work, we systematically validate this claim in both single- and multilingual settings by constructing training datasets with a controlled proportions of contextually relevant examples. We demonstrate a …

EvoCAD: Evolutionary CAD Code Generation with Vision Language Models
Tobias Preintner, Weixuan Yuan, Adrian K\"onig, Thomas B\"ack, Elena Raponi, Niki van Stein
https://arxiv.org/abs/2510.11631

EvoCAD: Evolutionary CAD Code Generation with Vision Language Models
Combining large language models with evolutionary computation algorithms represents a promising research direction leveraging the remarkable generative and in-context learning capabilities of LLMs with the strengths of evolutionary algorithms. In this work, we present EvoCAD, a method for generating computer-aided design (CAD) objects through their symbolic representations using vision language models and evolutionary optimization. Our method samples multiple CAD objects, which are then optimiz…

OpenAI says the cyber capabilities of its frontier AI models are accelerating and warns that upcoming models are likely to pose a "high" risk (Ina Fried/Axios)
https://www.axios.com/2025/12/10/openai-new-models-cybersecurity-risks

Exclusive: New OpenAI models likely pose "high" cybersecurity risk, company says
OpenAI says it's stepping up efforts to protect against attacks.

An End-to-End Differentiable, Graph Neural Network-Embedded Pore Network Model for Permeability Prediction
Qingqi Zhao, Heng Xiao
https://arxiv.org/abs/2509.13841 https://

An End-to-End Differentiable, Graph Neural Network-Embedded Pore Network Model for Permeability Prediction
Accurate prediction of permeability in porous media is essential for modeling subsurface flow. While pure data-driven models offer computational efficiency, they often lack generalization across scales and do not incorporate explicit physical constraints. Pore network models (PNMs), on the other hand, are physics-based and efficient but rely on idealized geometric assumptions to estimate pore-scale hydraulic conductance, limiting their accuracy in complex structures. To overcome these limitatio…

Improving Context Fidelity via Native Retrieval-Augmented Reasoning
Suyuchen Wang, Jinlin Wang, Xinyu Wang, Shiqi Li, Xiangru Tang, Sirui Hong, Xiao-Wen Chang, Chenglin Wu, Bang Liu
https://arxiv.org/abs/2509.13683

Improving Context Fidelity via Native Retrieval-Augmented Reasoning
Large language models (LLMs) often struggle with context fidelity, producing inconsistent answers when responding to questions based on provided information. Existing approaches either rely on expensive supervised fine-tuning to generate evidence post-answer or train models to perform web searches without necessarily improving utilization of the given context. We propose CARE, a novel native retrieval-augmented reasoning framework that teaches LLMs to explicitly integrate in-context evidence wi…

ThinkPilot: Steering Reasoning Models via Automated Think-prefixes Optimization
Sunzhu Li, Zhiyu Lin, Shuling Yang, Jiale Zhao, Wei Chen
https://arxiv.org/abs/2510.12063 https:/…

ThinkPilot: Steering Reasoning Models via Automated Think-prefixes Optimization
Large Reasoning Models (LRMs) are powerful, but they still suffer from inefficient and off-target reasoning. Currently, training-free methods are limited to either rigid heuristics or descriptive, non-actionable analyses. In this paper, we introduce ThinkPilot, a training-free framework that automatically optimizes LRMs reasoning. It uses an evolutionary process to generate think-prefixes, which are instructions that evolve driven by a taxonomy of reasoning behaviors to guide models toward supe…

Hybrid Quantum-Classical Neural Networks for Few-Shot Credit Risk Assessment
Zheng-an Wang, Yanbo J. Wang, Jiachi Zhang, Qi Xu, Yilun Zhao, Jintao Li, Yipeng Zhang, Bo Yang, Xinkai Gao, Xiaofeng Cao, Kai Xu, Pengpeng Hao, Xuan Yang, Heng Fan
https://arxiv.org/abs/2509.13818

Hybrid Quantum-Classical Neural Networks for Few-Shot Credit Risk Assessment
Quantum Machine Learning (QML) offers a new paradigm for addressing complex financial problems intractable for classical methods. This work specifically tackles the challenge of few-shot credit risk assessment, a critical issue in inclusive finance where data scarcity and imbalance limit the effectiveness of conventional models. To address this, we design and implement a novel hybrid quantum-classical workflow. The methodology first employs an ensemble of classical machine learning models (Logi…

Teaching According to Talents! Instruction Tuning LLMs with Competence-Aware Curriculum Learning
Yangning Li, Tingwei Lu, Yinghui Li, Yankai Chen, Wei-Chieh Huang, Wenhao Jiang, Hui Wang, Hai-Tao Zheng, Philip S. Yu
https://arxiv.org/abs/2509.13790

Teaching According to Talents! Instruction Tuning LLMs with Competence-Aware Curriculum Learning
Efficient instruction tuning aims to enhance the ultimate performance of large language models (LLMs) trained on a given instruction dataset. Curriculum learning as a typical data organization strategy has shown preliminary effectiveness in instruction tuning. However, current curriculum tuning methods suffer from the curriculum rigidity, since they rely solely on static heuristic difficulty metrics. These methods fail to adapt to the evolving capabilities of models during training, resulting i…

Multitask finetuning and acceleration of chemical pretrained models for small molecule drug property prediction
Matthew Adrian, Yunsie Chung, Kevin Boyd, Saee Paliwal, Srimukh Prasad Veccham, Alan C. Cheng
https://arxiv.org/abs/2510.12719

Multitask finetuning and acceleration of chemical pretrained models for small molecule drug property prediction
Chemical pretrained models, sometimes referred to as foundation models, are receiving considerable interest for drug discovery applications. The general chemical knowledge extracted from self-supervised training has the potential to improve predictions for critical drug discovery endpoints, including on-target potency and ADMET properties. Multi-task learning has previously been successfully leveraged to improve predictive models. Here, we show that enabling multitasking in finetuning of chemic…

DiffEM: Learning from Corrupted Data with Diffusion Models via Expectation Maximization
Danial Hosseintabar, Fan Chen, Giannis Daras, Antonio Torralba, Constantinos Daskalakis
https://arxiv.org/abs/2510.12691

DiffEM: Learning from Corrupted Data with Diffusion Models via Expectation Maximization
Diffusion models have emerged as powerful generative priors for high-dimensional inverse problems, yet learning them when only corrupted or noisy observations are available remains challenging. In this work, we propose a new method for training diffusion models with Expectation-Maximization (EM) from corrupted data. Our proposed method, DiffEM, utilizes conditional diffusion models to reconstruct clean data from observations in the E-step, and then uses the reconstructed data to refine the cond…

Mitigating the Noise Shift for Denoising Generative Models via Noise Awareness Guidance
Jincheng Zhong, Boyuan Jiang, Xin Tao, Pengfei Wan, Kun Gai, Mingsheng Long
https://arxiv.org/abs/2510.12497

Mitigating the Noise Shift for Denoising Generative Models via Noise Awareness Guidance
Existing denoising generative models rely on solving discretized reverse-time SDEs or ODEs. In this paper, we identify a long-overlooked yet pervasive issue in this family of models: a misalignment between the pre-defined noise level and the actual noise level encoded in intermediate states during sampling. We refer to this misalignment as noise shift. Through empirical analysis, we demonstrate that noise shift is widespread in modern diffusion models and exhibits a systematic bias, leading to …

Representation-Based Exploration for Language Models: From Test-Time to Post-Training
Jens Tuyls, Dylan J. Foster, Akshay Krishnamurthy, Jordan T. Ash
https://arxiv.org/abs/2510.11686

Representation-Based Exploration for Language Models: From Test-Time to Post-Training
Reinforcement learning (RL) promises to expand the capabilities of language models, but it is unclear if current RL techniques promote the discovery of novel behaviors, or simply sharpen those already present in the base model. In this paper, we investigate the value of deliberate exploration -- explicitly incentivizing the model to discover novel and diverse behaviors -- and aim to understand how the knowledge in pre-trained models can guide this search. Our main finding is that exploration wi…