• Bigger Teachers Do Not Guarantee Better Distillation Feedback. Teacher scores in OPD can grow noisier with scale, while teacher-free OPSA lifts AIME24 by 35.41 points.

• PaperGym Turns Research-Plan Review Into an Auditable Reward Environment. Separating problem and rubric sources cuts rubric leakage to 3.7%.

• CAST Uses Action-Level Critiques to Reduce Costly Tool-Calling Errors. It beats GPT-OSS-120B by over 10% on consecutive retail-task passes and transfers to telemedicine.

• NoisEasier Fixes Video Object Relationships Without Fine-Tuning. Test-time diffusion-noise optimization improves difficult dimensions such as attribute binding and object interaction by over 10% on average.

Also Notable

• One-Stage Pixel Diffusion Meets Linear Attention. Lapis combines both designs to balance generative depth quality with deployment efficiency.
• Why Test-Time Scaling Fails on Geometry. The study examines visual perception and multi-step symbolic deduction in plane geometry.
• Multimodal RAG Moves Beyond Single Images. Doc-REFRAG tests retrieval-augmented generation against realistic, multi-image document collections.
• Korean Jamo Errors Expose Language-Specific Weaknesses. The study measures and mitigates spelling perturbations that standard character-level tests miss.
• Representation Guidance Restores Safety After Fine-Tuning. Cross-family guidance controls internal representations at inference time without retraining specialized models.
• Saturated MCQA Benchmarks Need Probability Audits. Response-probability distributions may reveal question quality better than another batch of harder items.
• Scene-Adaptive Tone Mapping Improves HDR Detection. Per-scene adjustments help existing low-dynamic-range detectors extract more information from HDR inputs.
• Self-Developed Skills Create New Attack Surfaces. EvoSkill Injection tests malicious skill injection against agents that generate and modify their own skills.
• Internal Feature Transfer Narrows Cross-Language Gaps. The method transfers features from high-resource languages to low-resource languages on semantically equivalent reasoning tasks.
• Confidence Alone Cannot Detect Visual Hallucinations. VisER also measures whether the model truly depends on the corresponding image evidence.
• Multitask RL Should Adapt Its Data Mix. PAC shifts training toward tasks that still provide learning value and away from mastered material.

Read the full edition →