2026
10,558 articles · 132 days · 10 categories · avg 80 per day
PaperFormalizer Turns LaTeX Papers into Verified Lean Code, with the Author in the Loop
Stanford researchers released PaperFormalizer, an agentic pipeline for Claude Code or Codex that finds gaps in proofs and formalizes a paper in Lean.
NVIDIA open-sources Flora: robot data labeling 125 times cheaper than GPT-6 Astra
NVIDIA has open-sourced Flora, the labeling engine for GR00T. It labels 200 000 hours of data a day on 512 GB300 and costs about 125 times less than GPT-6 Astra.
G-Retriever's Graph Token Turns Out to Be a Dud: Cosine Similarity of 0.985–0.998
Researchers at Sapienza show that G-Retriever's "graph token" is nearly identical for any input, and a constant with no access to the graph matches the method's…
Anthropic: Claude Designed 318 484 Proteins in 64 Hours, and 17 Passed Lab Validation
CAS Institute of Software and Kuaishou Build First Benchmark for LLM User Simulation
RSIGame: a 27B model combo plays its own games and outperforms GPT-5.5
DecisionPivot Tests 40 Solver Models: Up to a 26-Point Gap at Equal Accuracy
S-Agent 2.0 teaches a coding agent to understand 3D and beats SpatialClaw by 7.8 points
A ready-made coding agent with a set of 3D tools lifted Qwen3.8-27B to 66.5 on MMSI-Bench and 82.3 on ReVSI. Gemini-3.8-Flash reached 74.1 and 90.5.
RiseHarness Teaches Terminal Agents to Fix Their Own Harness Without Fine-Tuning
RiseHarness iteratively improves the harness of terminal agents with frozen weights. It compares successful and failed runs of the same task and validates each…
EvaOmni Benchmark Tests 24 Omni Models: Leader Scores 79.4% Accuracy, but Only 35.1% on Grounding
Video-Point Model Predicts Multi-Camera Video and 3D Point Trajectories
Recursive Self-Improvement and Fable 5.1: Why the AI Race Moved Past AGI Debates Into Applied Orchestration
Vercel Releases durable-harness for Fault-Tolerant AI Agent Execution in the Cloud
SplatLab writes its own 3D Gaussian Splatting papers and beats baselines by up to +3.78 dB
An autonomous system reproduces a 3DGS paper, proposes an extension, trains models, and writes a manuscript. PSNR gains over baselines range from +0.18 to +3.78 dB.
Survey: Robots Are Learning to Improve Their Own Learning, but Evidence Is Still Missing
A new survey carries recursive self-improvement from code into the physical world and names four conditions under which robot gains accumulate instead of vanishing.
ToM2Vec Teaches LLM Agents to Model Their Partners in Latent Tokens
ToM2Vec builds a partner's theory of mind in continuous tokens and wins in Hanabi, Overcooked-AI and negotiation, including games with an unfamiliar GPT-5.4.
SpatialBrain Teaches a Frozen VLM to Accumulate Experience, Winning on 4 of 5 Benchmarks
DONIC teaches one policy to control two hands: 86.79% success without fine-tuning
TraceForge trains models on "their own thoughts": +7.1 points on MBPP+
Researchers Propose PAPAV Framework: Only 3 of 64 Benchmarks Test Verification
NUS Teaches a Model to Find What It Described: gIoU Rises From 57.5% to 68.2%
Generalized CycleGRPO links describing a region and finding it from text in a single RL loop, with no reference captions and no LLM judge. On GroundingSuite, gIoU…
TenetBench: 11 of 14 LLM Configurations Violate Professional Ethics Codes
A new Russian-language benchmark tests models against specific clauses of professional ethics codes. 11 of 14 configurations produce conflicting content on most…
LLM2Cert verifies 81 of 105 stability and safety certificates
Peking University speeds up video loading for AI by 1.45x
CausalAlign cures video models' "forgetfulness": geometry errors drop 20%
LEGO-SCD learns to detect changes in 3D scenes at 17.2 FPS without manual labeling
Six Teams from UCSB, Stanford and J.P. Morgan Release an Agent Memory Benchmark
WorldMemArena tests how multimodal agents remember and use what they have seen in real interactions. The six-organization paper has been accepted to NeurIPS 2026.
九章云极 Alaya-DFlash2 speeds up GLM-5.3 decoding by 3.41x
Chinese cloud provider 九章云极 unveiled Alaya-DFlash2: speculative decoding speeds up GLM-5.3 by 3.41x on long inputs, such as 32k-token documents.
MemT 1.0 builds memory into the model and aims at AI self-improvement
MemT 1.0 is out: memory is meant to be a built-in model capability, not an external add-on. A story about DeepSeek and an exhausted context sparked the discussion.
Anhui Bets on Humanoids: A Robotics Park in Hefei and a 15.1 Billion Yuan Project
Lovable Launches Auto-Deletion of "Dead" Apps: 60 Days of Silence, Then the Trash
Jiqizhixin Invites NeurIPS 2026 Authors to an Offline Meetup in Beijing
