Tootfinder

GPT-5.2: Neues KI-Modell von OpenAI soll Büroarbeiten besser unterstützen
Nur einen Monat nach GPT-5.1 kommt ein neues KI-Modell der ChatGPT-Entwickler. GPT-5.2 soll bessere Tabellen, Präsentationen und Code produzieren können.

OpenAI says GPT-5.2 Thinking hallucinates less than GPT-5.1 and has improved reliability for agentic AI needs; pre-release testers include Notion, Box, Shopify (Hayden Field/The Verge)
https://www.theverge.com/ai-artificial-intelligence/842529/open…

OpenAI calls GPT-5.2 the best model yet for professionals
OpenAI has launched GPT-5.2, which it calls its best AI model yet for professional use.

GPT-5.2 models match GPT-5 and 5.1 with a 400K context window and 128K max output tokens, but have a newer knowledge cutoff of Aug. 31, 2025 vs. Sept. 30, 2024 (Simon Willison/Simon Willison's Newsletter)
https://simonw.substack.com/p/gpt-52-and-useful-patterns-for-…

GPT 5.2 and useful patterns for building HTML tools
Plus a YouTube video, a podcast appearance and more

Single layer tiny Co$^4$ outpaces GPT-2 and GPT-BERT
Noor Ul Zain, Mohsin Raza, Ahsan Adeel
https://arxiv.org/abs/2510.08404 https://arxiv.org/pdf/2510.084…

Single layer tiny Co$^4$ outpaces GPT-2 and GPT-BERT
We show that a tiny Co$^4$ machine(Adeel,2025) with a single layer, two heads, and 8M parameters, operating at an approximate cost of $O(N)$ (where $N$ is the number of input tokens), outpaces the BabyLM Challenge baselines GPT-2 (124M, 12 layers, $O(N^2))$ and GPT-BERT (30M, 12 layers, $O(N^2))$ in just two epochs, while both are trained for ten. Co$^4$ achieves orders-of-magnitude greater training efficiency on 10M tokens, demonstrating highly sample efficient pretraining. Using the BabyLM ch…

i got up at 6 a.m. to wait in line

OpenAI launches GPT-5.2, its "best model yet," in Instant, Thinking, and Pro variants, with significant improvements in writing, coding, and reasoning (Maxwell Zeff/Wired)
https://www.wired.com/story/openai-gpt-launch-gemini-code-red/

OpenAI Launches GPT-5.2 as It Navigates ‘Code Red’
The ChatGPT-maker is releasing its “best model yet” as it faces new pressures from Google and other AI competitors.

Freitag: Kritik an eID-Karte wegen Geldwäsche, neues OpenAI-Modell als Bürohilfe
eID-Karte zu einfach zu ergaunern GPT-5.2 für Profi-Nutzer Disney gegen Google-KI wegen Copyright Kritik an EU wegen VMware Roboter-Bewegungen erklärt

Weight Initialization and Variance Dynamics in Deep Neural Networks and Large Language Models
Yankun Han
https://arxiv.org/abs/2510.09423 https://arxiv.org…

Weight Initialization and Variance Dynamics in Deep Neural Networks and Large Language Models
Weight initialization governs signal propagation and gradient flow at the start of training. This paper offers a theory-grounded and empirically validated study across two regimes: compact ReLU multilayer perceptrons and GPT-2-style transformers. First, a logarithmic sweep of the initial standard deviation maps vanishing and exploding regimes and identifies a broad stability band with standard deviations between 1e-2 and 1e-1. Second, a controlled comparison shows that Kaiming (fan-in) initiali…

XplaiNLP at CheckThat! 2025: Multilingual Subjectivity Detection with Finetuned Transformers and Prompt-Based Inference with Large Language Models
Ariana Sahitaj, Jiaao Li, Pia Wenzel Neves, Fedor Splitt, Premtim Sahitaj, Charlott Jakob, Veronika Solopova, Vera Schmitt
https://arxiv.org/abs/2509.12130

XplaiNLP at CheckThat! 2025: Multilingual Subjectivity Detection with Finetuned Transformers and Prompt-Based Inference with Large Language Models
This notebook reports the XplaiNLP submission to the CheckThat! 2025 shared task on multilingual subjectivity detection. We evaluate two approaches: (1) supervised fine-tuning of transformer encoders, EuroBERT, XLM-RoBERTa, and German-BERT, on monolingual and machine-translated training data; and (2) zero-shot prompting using two LLMs: o3-mini for Annotation (rule-based labelling) and gpt-4.1-mini for DoubleDown (contrastive rewriting) and Perspective (comparative reasoning). The Annotation App…

[Thread] GPT-5.2 is now available in the API, priced at $1.75/1M input and $14/1M output tokens; GPT-5.2 Pro is priced at $21/1M input and $168/1M output tokens (@openaidevs)
https://x.com/openaidevs/status/1999184802755354954

OpenAI Developers (@OpenAIDevs) on X
GPT-5.2 is now available in the API—our most advanced frontier model for long-running agents. https://t.co/J5i4czpKsL

CapGeo: A Caption-Assisted Approach to Geometric Reasoning
Yuying Li, Siyi Qian, Hao Liang, Leqi Zheng, Ruichuan An, Yongzhen Guo, Wentao Zhang
https://arxiv.org/abs/2510.09302 …

CapGeo: A Caption-Assisted Approach to Geometric Reasoning
Geometric reasoning remains a core challenge for Multimodal Large Language Models (MLLMs). Even the most advanced closed-source systems, such as GPT-O3 and Gemini-2.5-Pro, still struggle to solve geometry problems reliably, despite exhibiting strong textual reasoning abilities on tasks like the International Mathematical Olympiad (IMO). This gap suggests that the bottleneck lies in understanding geometric diagrams rather than reasoning itself. Since geometric figures can often be faithfully des…

Growing Perspectives: Modelling Embodied Perspective Taking and Inner Narrative Development Using Large Language Models
Sabrina Patania, Luca Annese, Anna Lambiase, Anita Pellegrini, Tom Foulsham, Azzurra Ruggeri, Silvia Rossi, Silvia Serino, Dimitri Ognibene
https://arxiv.org/abs/2509.11868

Growing Perspectives: Modelling Embodied Perspective Taking and Inner Narrative Development Using Large Language Models
Language and embodied perspective taking are essential for human collaboration, yet few computational models address both simultaneously. This work investigates the PerspAct system [1], which integrates the ReAct (Reason and Act) paradigm with Large Language Models (LLMs) to simulate developmental stages of perspective taking, grounded in Selman's theory [2]. Using an extended director task, we evaluate GPT's ability to generate internal narratives aligned with specified developmental stages, a…

Baidu unveils Ernie 5.0, an AI model to process and generate text, images, audio, and video, claiming it beats GPT-5-High and Gemini 2.5 Pro on some benchmarks (Carl Franzen/VentureBeat)
https://venturebeat.com/ai/baidu-unveils-proprietary-ern…

Kimi K2 is another Deepseek moment it seems, only not everybody is noticing it yet. It will be interesting to see what the stock market will do on monday.
#AI #KimiK2

Anthropic open sources a method to score AI model political evenhandedness; Gemini 2.5 Pro got 97%, Grok 4 96%, Claude Opus 4.1 95%, GPT-5 89%, and Llama 4 66% (Ina Fried/Axios)
https://www.axios.com/2025/11/13/anthropic-bot-bias-data

Anthropic tries to measure bot bias
There's no consensus over what constitutes political bias in AI systems.

OpenAI says GPT‑5.2 Thinking beats or ties industry professionals on 70.9% of GDPval knowledge work tasks, delivering outputs at >11x the speed and <1% the cost (OpenAI)
https://openai.com/index/introducing-gpt-5-2

Curious that whenever someone shows me “the cool #AI flow” they built that’s supposed to be impressive, the conversation goes the same way:
Stage 1: “But you don’t understand. You don’t like AI because you haven’t used it right. Let me show you how much you can do it with.”
Stage 2: “Here are the steps in the flow and the instructions I feed to this agent / custom GPT / Claude project. I tell it to do X, reference document Y, and aim for Z.”
Stage 3: “Now, let me show you the results it gives.”
*Writes task, presses to run the prompt.*
Stage 4: “Umm sorry it’s taking a while. It’s fast but not instant. And by the way, the prompt isn’t perfect, you can definitely make it better. I just threw this together real quick the other day. It makes some mistakes, but it’s really good.”
Stage 5: “Uuuuuuh actually don’t look at the output.” *scrolls or stops screen share or pulls device away.*
“You know it’s already doing so well, if I do more prompt engineering it will get really good but I need to give it better instructions. And it ran just fine last night, I don’t know what’s up with it. And this is a cheap model, if we use another model it will be better.”
Stage 6: “You know, you really shouldn’t judge this so much. The technology will improve, it will get there sooner than you know and then you’ll regret not trying it sooner.”
So curious that this keeps happening 🤷‍♀️
#LLMs #work #tech #AIBubble

OpenAI releases an invitation-only Sora app on iOS, powered by Sora 2, to let people create and share AI-generated videos of themselves and their friends (Ina Fried/Axios)
https://www.axios.com/2025/09/30/openai-sora-app-social-ai

OpenAI's Sora app lets friends swap AI video cameos
The Chat-GPT maker joins Meta in betting that AI-generated video is the new social media.

MERCER ISLAND SCHOOL BOARD
Wow, you really do have to watch the downballot races. Mercer Island School Board has two (2) candidates (O'Callahan is and Gaspar) that are *both* software CTOs touting their "AI" credentials. Gaspar explicitly wants "free AI classes".
Here's a hint: The only "AI classes" that kids need are ones that teach them how to TURN ALL OF THAT SHIT OFF, and learn to think and write in their own words, not Chat-GPT'…

🧾 Multi-Modal Vision vs. Text-Based Parsing: Benchmarking LLM Strategies for Invoice Processing
#software

Multi-Modal Vision vs. Text-Based Parsing: Benchmarking LLM Strategies for Invoice Processing
This paper benchmarks eight multi-modal large language models from three families (GPT-5, Gemini 2.5, and open-source Gemma 3) on three diverse openly available invoice document datasets using zero-shot prompting. We compare two processing strategies: direct image processing using multi-modal capabilities and a structured parsing approach converting documents to markdown first. Results show native image processing generally outperforms structured approaches, with performance varying across mode…

Companies are updating insider trading policies to cover prediction markets; Kalshi and others are pushing for federal oversight, including of insider trading (Rocket Drew/The Information)
https://www.theinformation.com/articles/polymark…

Polymarket Bets on OpenAI, Google Raise Insider Trading Suspicions
Over the past week, a handful of accounts on Polymarket, the predictions site, bet OpenAI would release a new large language model by December 13. On Thursday December 11, OpenAI released GPT-5.2, and four of these accounts together made over $13,000, according to the trades displayed on their ...

Resource-Efficient Fine-Tuning of LLaMA-3.2-3B for Medical Chain-of-Thought Reasoning
Imran Mansha
https://arxiv.org/abs/2510.05003 https://arxiv.org/pdf/2…

Resource-Efficient Fine-Tuning of LLaMA-3.2-3B for Medical Chain-of-Thought Reasoning
Large Language Models (LLMs) such as GPT-4 and LLaMA have demonstrated remarkable reasoning abilities but require significant computational resources for fine-tuning. This paper presents a resource-efficient fine-tuning approach for LLaMA-3.2-3B to enhance medical chain-of-thought reasoning while operating under constrained GPU and memory settings. Using parameter-efficient tuning techniques such as LoRA and QLoRA, we adapt the base model on publicly available medical reasoning datasets. The mo…

A Deep Learning Pipeline for Epilepsy Genomic Analysis Using GPT-2 XL and NVIDIA H100
Muhammad Omer Latif, Hayat Ullah, Muhammad Ali Shafique, Zhihua Dong
https://arxiv.org/abs/2510.00392

A Deep Learning Pipeline for Epilepsy Genomic Analysis Using GPT-2 XL and NVIDIA H100
Epilepsy is a chronic neurological condition characterized by recurrent seizures, with global prevalence estimated at 50 million people worldwide. While progress in high-throughput sequencing has allowed for broad-based transcriptomic profiling of brain tissues, the deciphering of these highly complex datasets remains one of the challenges. To address this issue, in this paper we propose a new analysis pipeline that integrates the power of deep learning strategies with GPU-acceleration computat…

OpenAI announces API updates, including GPT-5 Pro, Sora 2 in preview, and gpt-realtime-mini, a voice model that is 70% cheaper than gpt-realtime (Rebecca Bellan/TechCrunch)
https://techcrunch.com/2025/10/06/openai-ramps-up-developer-push…

OpenAI ramps up developer push with more powerful models in its API  | TechCrunch
The updates were part of a series of announcements geared towards wooing developers to OpenAI’s ecosystem, including the launch of an agent-building tool and the ability to build apps in ChatGPT. 

Replaced article(s) found for cs.AI. https://arxiv.org/list/cs.AI/new
[2/6]:
- Understanding AI Evaluation Patterns: How Different GPT Models Assess Vision-Language Descriptions
Sajjad Abdoli, Rudi Cilibrasi, Rima Al-Shikh

Evaluating LLM-Generated Legal Explanations for Regulatory Compliance in Social Media Influencer Marketing
Haoyang Gui, Thales Bertaglia, Taylor Annabell, Catalina Goanta, Tjomme Dooper, Gerasimos Spanakis
https://arxiv.org/abs/2510.08111

Evaluating LLM-Generated Legal Explanations for Regulatory Compliance in Social Media Influencer Marketing
The rise of influencer marketing has blurred boundaries between organic content and sponsored content, making the enforcement of legal rules relating to transparency challenging. Effective regulation requires applying legal knowledge with a clear purpose and reason, yet current detection methods of undisclosed sponsored content generally lack legal grounding or operate as opaque "black boxes". Using 1,143 Instagram posts, we compare gpt-5-nano and gemini-2.5-flash-lite under three prompting str…

OpenAI releases an invitation-only Sora app on iOS, powered by Sora 2, to let people create and share AI-generated videos of themselves and their friends (Ina Fried/Axios)
https://www.axios.com/2025/09/30/openai-sora-app-social-ai

OpenAI's Sora app lets friends swap AI video cameos
The Chat-GPT maker joins Meta in betting that AI-generated video is the new social media.

OpenAI releases gpt-oss-safeguard, its open-weight reasoning models for safety classification tasks, available in 120B and 20B parameters, under Apache 2.0 (OpenAI)
https://openai.com/index/introducing-gpt-oss-safeguard/

Alibaba Technical Report: Qwen3-VL beats GPT-5 and Gemini 2.5 Pro on visual tasks and has 100% accuracy on "needle-in-a-haystack" tests for 30-minute videos (Jonathan Kemper/The Decoder)
https://the-decoder.com/qwen3-vl-can-scan-two-hour-…

Qwen3-VL can scan two-hour videos and pinpoint nearly every detail
A few months after launching Qwen3-VL, Alibaba has released a detailed technical report on the open multimodal model. The data shows the system excels at image-based math tasks and can analyze hours of video footage.

Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct
Haoyang Zheng, Xinyang Liu, Cindy Xiangrui Kong, Nan Jiang, Zheyuan Hu, Weijian Luo, Wei Deng, Guang Lin
https://arxiv.org/abs/2509.25035

Ultra-Fast Language Generation via Discrete Diffusion Divergence Instruct
Fast generation of language texts is the holy grail that people pursue in the AI era. In this work, we introduced Discrete Diffusion Divergence Instruct (DiDi-Instruct), a training-based method that leads to fast language generation models by initializing from a pre-trained (masked) discrete diffusion language model (dLLM). The resulting DiDi-Instruct model outperforms the dLLM counterparts and the GPT-2 baseline with 64x acceleration. In the theoretical part of the paper, we build the foundati…

Gemini 3 Pro is priced at $2-$4 per 1M input tokens and $12-$18 per 1M output tokens, cheaper than Claude Sonnet 4.5 but more expensive than GPT-5.1 (Simon Willison/Simon Willison's Weblog)
https://simonwillison.net/2025/Nov/18/gemini-3/

Trying out Gemini 3 Pro with audio transcription and a new pelican benchmark
Google released Gemini 3 Pro today. Here’s the announcement from Sundar Pichai, Demis Hassabis, and Koray Kavukcuoglu, their developer blog announcement from Logan Kilpatrick, the Gemini 3 Pro Model Card, …

Investigating Bias: A Multilingual Pipeline for Generating, Solving, and Evaluating Math Problems with LLMs
Mariam Mahran, Katharina Simbeck
https://arxiv.org/abs/2509.17701 htt…

Investigating Bias: A Multilingual Pipeline for Generating, Solving, and Evaluating Math Problems with LLMs
Large Language Models (LLMs) are increasingly used for educational support, yet their response quality varies depending on the language of interaction. This paper presents an automated multilingual pipeline for generating, solving, and evaluating math problems aligned with the German K-10 curriculum. We generated 628 math exercises and translated them into English, German, and Arabic. Three commercial LLMs (GPT-4o-mini, Gemini 2.5 Flash, and Qwen-plus) were prompted to produce step-by-step solu…

A Comparative Evaluation of Large Language Models for Persian Sentiment Analysis and Emotion Detection in Social Media Texts
Kian Tohidi, Kia Dashtipour, Simone Rebora, Sevda Pourfaramarz
https://arxiv.org/abs/2509.14922

A Comparative Evaluation of Large Language Models for Persian Sentiment Analysis and Emotion Detection in Social Media Texts
This study presents a comprehensive comparative evaluation of four state-of-the-art Large Language Models (LLMs)--Claude 3.7 Sonnet, DeepSeek-V3, Gemini 2.0 Flash, and GPT-4o--for sentiment analysis and emotion detection in Persian social media texts. Comparative analysis among LLMs has witnessed a significant rise in recent years, however, most of these analyses have been conducted on English language tasks, creating gaps in understanding cross-linguistic performance patterns. This research ad…