#Research
Curated coverage on «Research»: releases, research, benchmarks, open-source.
GlobeBench: The Best LLM Correctly Predicts Environment Responses in Only 49.57% of Tasks
The 1632-task GlobeBench found that the best of 20 models, GPT-6 Astra, simulates the environment's response correctly in only 49.57% of cases. No model solved 631…
A Frozen 16 MB Vocabulary Matches a Trained One in Small LLMs
Researchers replaced 168M embedding parameters in a Gemma-like model with a frozen 16 MB table and did not lose to the trainable version on three of four benchmarks.
LIBERO-Industry tests 5 robot policies on 216 factory tasks
The new LIBERO-Industry benchmark adds a conveyor, narrow action windows, and tight tolerances. Five generalist policies lose performance as conditions get harder.
Tsinghua Unveils SRHarness and MechBench: AI Agents That Look for Formulas and Mechanisms
Tsinghua Unveils Jev-LDE: Edit ICL Examples Once, Boost 1-Shot Accuracy
Recursia Beats Nvidia ASPIRE by 10 Points on Long-Horizon Robot Tasks
ETH and Roche Train AI to Predict Where C–H Borylation Will Occur on a Molecule
PoS Method Catches AI Agents Stuck in Loops of False Belief: 36 Repeats in ALFWorld
An ALFWorld agent forgot to heat a cup and checked the cabinet 36 times in a row until its action budget ran out. A new paper calls this Belief Trapping and…
Ego Data Gets a "Full Body": A New Scaling Path for World Models
Robot data collection is shifting from expensive platforms to lightweight formats. Jiqizhixin examines the idea of extending ego recordings with whole-body motion…
ReaDy speeds up hybrid LLM decoding 3.16x on a single B200
DAGFormer Plans Connections Between Layers and Cuts Perplexity by 17–25%
Researchers Find a 0.4–1.1% Window in Pythia Pretraining Where the Readout Reorganizes
semCER Weights OCR Errors by Severity: 61M LLM-Judge Ratings
RTAformer Teaches a Model to Notice When Relationships Between Objects Change in Video
Researchers propose RTAformer, a one-stage transformer for video scene graphs. It models object persistence and relationship changes separately, and improves recall…
MolLean Makes LLMs Check Their Chemistry: Wins on 12 of 12 Tasks
MolLean, a new Lean-based language, verifies every step of a model's chemical reasoning. On Qwen3.6-27B it beat plain chain-of-thought on all 12 tasks across 8,800…