Tootfinder

North Korea on Thursday displayed apparent progress in the construction of a
nuclear-powered submarine,
with state media photos showing a largely completed hull,
as leader Kim Jong Un condemned rival South Korea’s push to acquire the technology.
North Korea’s official Korean Central News Agency said Kim visited a shipyard to inspect the construction of what the North describes as
an 8,700-ton-class nuclear-propelled submarine,
which the leader has called a…

North Korea shows progress on nuclear-powered submarine
North Korea displayed apparent progress in the construction of a nuclear-powered submarine, with state media photos showing a largely completed hull as leader Kim Jong Un condemned rival South Korea’s push to acquire the technology. North Korea’s official Korean Central News Agency on Thursday said Kim had visited a shipyard to inspect the construction of what the North describes as an 8,700-ton-class nuclear-propelled submarine, which he has previously called a crucial step in the moderniz…

What did countries agree to at COP30? #environment

What did countries agree to at COP30?
The COP30 climate summit has drawn to a close after two weeks in the Amazonian city of Belem where protests, street marches and even a fire caused unexpected moments of drama.

OLX Group, an online marketplace owned by Prosus, agrees to acquire La Centrale, a French auto classifieds platform, for $1.3B in cash, expected to close in Q4 (Loni Prinsloo/Bloomberg)
https://www.bloomberg.com/news/articles/20

UserRL: Training Interactive User-Centric Agent via Reinforcement Learning
Cheng Qian, Zuxin Liu, Akshara Prabhakar, Jielin Qiu, Zhiwei Liu, Haolin Chen, Shirley Kokane, Heng Ji, Weiran Yao, Shelby Heinecke, Silvio Savarese, Caiming Xiong, Huan Wang
https://arxiv.org/abs/2509.19736

UserRL: Training Interactive User-Centric Agent via Reinforcement Learning
Reinforcement learning (RL) has shown promise in training agentic models that move beyond static benchmarks to engage in dynamic, multi-turn interactions. Yet, the ultimate value of such agents lies in their ability to assist users, a setting where diversity and dynamics of user interaction pose challenges. In this work, we propose UserRL, a unified framework for training and evaluating user-centric abilities through standardized gym environments paired with simulated users. We systematically v…

Explaining Fine Tuned LLMs via Counterfactuals A Knowledge Graph Driven Framework
Yucheng Wang, Ziyang Chen, Md Faisal Kabir
https://arxiv.org/abs/2509.21241 https://

Explaining Fine Tuned LLMs via Counterfactuals A Knowledge Graph Driven Framework
The widespread adoption of Low-Rank Adaptation (LoRA) has enabled large language models (LLMs) to acquire domain-specific knowledge with remarkable efficiency. However, understanding how such a fine-tuning mechanism alters a model's structural reasoning and semantic behavior remains an open challenge. This work introduces a novel framework that explains fine-tuned LLMs via counterfactuals grounded in knowledge graphs. Specifically, we construct BioToolKG, a domain-specific heterogeneous knowled…

Agora que estou finamente a tentar controlar o caos que é o meu email chego Š conclusão de que aí uns 90% dos emails que recebo do meu banco são a alertar para fraudes.

Evaluating the Evaluators: Metrics for Compositional Text-to-Image Generation
Seyed Amir Kasaei, Ali Aghayari, Arash Marioriyad, Niki Sepasian, MohammadAmin Fazli, Mahdieh Soleymani Baghshah, Mohammad Hossein Rohban
https://arxiv.org/abs/2509.21227

Evaluating the Evaluators: Metrics for Compositional Text-to-Image Generation
Text-image generation has advanced rapidly, but assessing whether outputs truly capture the objects, attributes, and relations described in prompts remains a central challenge. Evaluation in this space relies heavily on automated metrics, yet these are often adopted by convention or popularity rather than validated against human judgment. Because evaluation and reported progress in the field depend directly on these metrics, it is critical to understand how well they reflect human preferences. …

Alphabet agrees to acquire clean energy developer Intersect for $4.75B in cash, and existing debt, as part of its push to dramatically expand AI data centers (Bloomberg)
https://www.bloomberg.com/news/articles/2025-12-22/alpha…

Alphabet to Buy Data Center Partner Intersect for $4.75 Billion
Alphabet Inc. has agreed to buy clean energy developer Intersect Power LLC for $4.75 billion in cash, plus its existing debt, marking one of the largest deals by the tech giant in its push to dramatically expand its data center footprint to power AI.

FINANCIAL TIMES
The US and Ukraine have drafted a new 19-point peace deal,
but left the most politically sensitive elements to be decided by the countries' presidents,
according to Ukrainian first deputy foreign minister Sergiy Kyslytsya.
Washington had previously put Kyiv under pressure to agree a 28-point proposal that had been developed by US and Russian officials and crossed several long-standing Ukrainian red lines.
Kyslytsya, who was in the room as par…

US and Ukraine draft new 19-point peace plan but defer biggest decisions
The most politically sensitive elements are left for Trump and Zelenskyy to discuss

The Canada Pension Plan Board and Australia's Goodman agree to invest $2.6B in data centers in Frankfurt, Amsterdam, and Paris, starting to build in June 2026 (Angus Whitley/Bloomberg)
https://www.bloomberg.com/news/articles/20