Post-training a language model on protein structure questions improves its performance on benchmarks that contain no proteins, no structural inputs, and no specialized modules. That is the central claim, and the paper…
CrewAI organizes multi-agent work around two distinct abstractions that can operate independently or together. Crews are teams of autonomous agents with defined roles, goals, and backstories that collaborate through…
MotorMind scaffolds a general VLM for zero-shot robot manipulation, hitting 66.7% on LIBERO-PRO vs 13.3% for prior methods - no task-specific training needed.
Read by the desk
↑111hf upvotes
RESEARCHSTEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization · arXiv · Oct 09
STEPQuant quantizes Delta-rule recurrent states spatially and temporally, hitting FP32-level accuracy at 6-bit and cutting total serving memory by up to 68.7%.
Read by the desk
↑108hf upvotes
RESEARCHScaling Trajectories for Complex Tasks through Recursive Self-Rewrite · arXiv · Oct 09
Long-WAM pairs autoregressive video pretraining with streaming encoding so robots can use up to 19.2 seconds of visual history under real-time control.
Read by the desk
↑97hf upvotes
REPOQingYunA/answer-me-with-html · GitHub · Oct 09
DN-MOPD rescales each specialist's feedback by its measured spread, fixing the imbalance that lets instruction-following dominate multi-teacher distillation.
Read by the desk
↑140hf upvotes
RESEARCHSelf-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence · arXiv · Sep 30
HexaAnything wraps VLA/WAM policies in code-represented task state so verified execution traces feed back as training data, beating direct VLA on unseen tasks.
Duplex-MPE is a benchmark of 2,000 scenarios testing whether a full-duplex speech assistant answers, stays silent, or stops in multi-party conversations.
Disaggregated quantization uses separate formats for prefill and decode, yielding 32+ point accuracy gains and 1.78x faster first-token latency on 27B models.
InternW0-Δ unifies visual dynamics, 4D geometry, and action generation in a Mixture-of-Transformers WAM trained on 20K+ hours of open robot and egocentric data.
WROP pairs a 1.5M-sample object-permanence corpus with a 300-question blind Elo exam; their 16B PWM-WROP ranks first among continuation video world models.
Read by the desk
↑193hf upvotes
RESEARCHAgent-Editing World Model: Rethinking World Modeling for LLM Agents · arXiv · Sep 26
AEWM edits noisy reasoning and actions in task history rather than predicting tool responses, improving agent scores by 3.2–6.7 points across six benchmarks.
Read by the desk
↑16hf upvotes
RESEARCHParts-of-Speech as Emergent Categories in SAE Latent Space · arXiv · Sep 26
A study finding that part-of-speech distinctions are recoverable from SAE activations via compact latent groups, not one-to-one latent-to-category mappings.
Read by the desk
↑10hf upvotes
RESEARCHQwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents · arXiv · Sep 26
RewardVerse uses generated rubrics as an intermediate step before scoring to reduce scalar drift in video reward models trained with a two-stage RL algorithm.
Read by the desk
↑22hf upvotes
RESEARCHSchrödinger's Code Repository: Have LLMs Learned SWE-bench or Memorized It? · arXiv · Sep 25
An evaluation framework that dynamically transforms test repositories to detect whether coding agents rely on memorized cues rather than genuine reasoning.
MemBodied adds fixed-size episodic memory to Vision-Language-Action models, achieving 7.81× the success rate of a stateless policy on memory-dependent tasks.