#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.
Cohere taught a 7B model to assess its confidence twice, cutting calibration error by 52.5%
Cohere's CALIBER method trains a reasoning model to estimate its odds before and after solving. On BigMathDigits, ECE fell from 0.059 to 0.028, with an AUROC of…
Claude Designed 17 Proteins Against Scorpion Toxins in 64 Hours
Anthropic's internal model generated 318 484 protein designs, and 211 were lab-tested. 17 bind scorpion neurotoxins; the best show nanomolar affinity.
250 subagents vs. 4 threads: why orchestration architecture matters more than a loud model name
Viral demos of new models like GPT-6 Astra are often built on third-party open-source libraries and partial checkpoint completion rather than actual task completion.
Dream Labs, from the NVIDIA DreamZero team, raises $90M and opens an API for robots
RLSplat teaches 3D Gaussian Splatting to economize: up to 4.1× faster and 9.6× less memory
Stratified Sparse Attention speeds up decoding 2.96x at 1M context
StyleStream 2.0 Cuts Voice Conversion Latency in Half, to 520 ms
Diffusion LLM Agents Learn to Whisper in Vectors: +34 Points on QASC
Researchers had diffusion language models exchange four hidden vectors instead of text. Accuracy on QASC rose to 80.8%, and a single late handoff reached 87.8%.
WorldStream Builds an Endless 3D World and Remembers Every Player Edit
WorldStream extends a playable 3D world beyond the map, executes language commands, and restores visited places with their edits. It plans on two levels…
A Team of AI Researchers Found a Linear Attention Design That Beat GDN2 at 1.3B
DiffResearch: A Diffusion LLM Makes Deep Research About 6x Faster
Of 1 598 RL Papers, 37 Made It Into Frameworks: An Audit of Seven Open-Source Stacks
Researchers Ran 120 CPT Configurations for Game Agents, From 0.8B to 27B
RoboJudge-9B Judges Robot Videos More Accurately Than Gemini: 0.719 Correlation With Humans
CUHK researchers built an 800-video annotated benchmark and trained a 9B judge for robot videos: 0.719 correlation with humans, 0.089 higher than Gemini-3.7-Flash.
DeCert Teaches AI Judges to Explain Their Verdicts: An 8B Model Scores 76.2 on RM-Bench
The DeCert framework trains generative reward models not just to pick the right answer but to justify it. Qwen3-8B scores 76.2 on RM-Bench, with up to +14.8 in…