Engineering
Language Models Can Control Their Own Attention
Researchers introduce Declarative Attention (DA), a protocol that has a language model explicitly declare in its chain-of-thought which parts of…
Engineering
S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?
S3Gym is a new interactive benchmark testing whether LLM agents can turn accumulated experience into genuine self-improvement, evaluating three coupled…
Engineering
RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests
A new study finds that real-world coding requests differ sharply from the curated GitHub issues used in SWE-bench-style benchmarks: 88%…
Engineering
Cliff: Learning Process Rewards from the First Mistake
Cliff is a new reward-shaping method for reinforcement learning with verifiable rewards (RLVR) that uses an off-the-shelf LLM to identify…
Engineering
Using Grounded Theory for Agent Behavior Analysis at Scale
Researchers adapt grounded theory, a decades-old qualitative research method from the social sciences, into AutoTraceGT, the first automated multi-agent pipeline…
Engineering
Environment Evolution for Terminal Agents
A new paper proposes environment evolution, a method that incrementally increases the difficulty of training environments for terminal agents off-policy…
Engineering
NeoMME: A Single-Tower Multimodal-Native Multilingual Foundation Encoder for Efficient Fine-Tuning and Inference
NeoMME introduces a family of compact 260M and 800M-parameter bidirectional encoders that process multilingual text and raw image patches within…
News
Z.ai Launches GLM-5.3-Flash, Its First Natively Multimodal GLM-5 Model
Z.ai released GLM-5.3-Flash, a natively multimodal mixture-of-experts model with 320 billion total parameters and 18 billion active parameters, published under…
News
xAI Expands Grok Bot Access to SuperGrok and Cursor Plans
xAI announced that Grok Bot, its always-on AI agent product, is now included with SuperGrok, Cursor Pro, and all Cursor…
News
Anthropic Says Claude Formalized a Complete Proof of Fermat’s Last Theorem
Anthropic researchers announced that Claude produced the first complete, computer-verified proof of Fermat’s Last Theorem in the Lean proof assistant,…
News
Google DeepMind Launches WeatherNext 3 With Hourly, High-Resolution AI Forecasts
Google DeepMind introduced WeatherNext 3, an AI weather forecasting model that delivers hourly predictions at five times higher resolution than…
News
OpenAI Unveils GPT-6 Astra, Its Most Intelligent and Aligned Model Yet
OpenAI unveiled GPT-6 Astra, described as “the world’s most intelligent and aligned model,” featuring significant improvements in computer use, software…