A new arXiv paper (2609.38817v1) introduces RADAR (Reasoning-state Analysis via Dynamic Attention Responses), a method that identifies a large reasoning model's current reasoning state in real time an…
A 2x2 randomized field experiment embedded in a public German Santa Claus telephone hotline found that persuasive framing raised the probability of a prosocial wish from 11.6% to 45.7%, while persona …
Researchers introduced EvoSteer, an online self-evolving graph orchestration method for LLM-based multi-agent systems, with code available at https://github.com/beita6969/evosteer. EvoSteer pairs Anch…
A new arXiv paper (2609.38670v1) introduces decision checkpoints that record observations and tool actions during inference to distinguish whether a scientific-search agent failed to encounter a targe…
A new arXiv paper (2609.38829v1) formalizes multi-path LLM reasoning as a diversity combining problem from wireless communications, showing that pairwise correlation of path correctness caps the effec…
Researchers introduced Code to Control, an approach that synthesizes Python controllers executing directly as policies, separating program structure from parameters so an LLM builds the controller str…
GraphCert, a post-training method from researchers behind the arXiv paper 2609.38798v1, bootstraps agentic graph reasoning by generating graph-grounded QA pairs and certifying supporting evidence into…
A working implementation of reinforcement learning for calibrated decisions (RLCD) for reasoning models, OpenJev-RLCD, matches or beats supervised fine-tuning, RFT/STaR and GRPO in accuracy and beats …
Researchers introduced Mamushi, a non-parametric framework for distributional unlearning that ranks forget examples using a probabilistic classifier whose Bayes-optimal logit equals the forget-to-reta…
A new arXiv paper (2609.38891v1) introduces Consistent Plan-Act (ConPAct), a method that detects and feeds back contradictions between a planner agent and an actor agent to form consistent state inter…
A new arXiv paper (2609.38962v1) proposes USteer, a training-free steering mechanism that reduces hallucination in large language model reasoning tasks by adjusting layer-wise activations during infer…
A new arXiv paper (2609.38964v1) reports that sequential multi-agent debate (MAD) among large language models suffers from a pronounced first-speaker bias, in which agents disproportionately shape the…
Researchers introduced C-STRIDE, an observation-driven AI digital twin that predicts basin-wide flood water depth from sparse stream-gauge records, terrain, and rainfall, extending forecasts up to one…
A new arXiv paper (arXiv:2609.39101v1) introduces Retrospective World Modeling, an agent learning paradigm that estimates the retrospective attribution distribution P(â_t|s_t, s_{t+1}) to identify the…
A new arXiv paper (2609.38958v1) reports that Chain-of-Thought reasoning improves long-context counting accuracy across twelve model comparison groups, with the largest gains at higher counts. The aut…
A study of five self-evolving skill methods and a one-shot skill across six benchmarks found that of 21 skills that improved on their training tasks, only 5 kept all of that improvement on held-out te…
OpenAI said it identified an attempt to extract protected reasoning from its AI models and linked part of the activity to China's Moonshot AI, CNBC reported. The disclosure ties a Chinese AI developer…
California Governor Gavin Newsom signed a state law banning AI 'robo bosses' from firing workers, making California the first U.S. state with such a law and reversing his earlier veto of the measure. …
Google released Gemini 4 Argon, its first frontier model in over seven months, which matches OpenAI's GPT-6 Astra in independent testing but trails Anthropic's Claude Opus 5.5, according to The Decode…
Google rolled out Gemini 4 Argon, which the company calls Alphabet's most advanced AI model yet, according to CNBC Technology. The model adds major improvements in coding, cybersecurity, and complex p…