• Agent failures split into two measurable error modes: locking onto one path (over-exploit) and wandering without direction (over-explore) can be separated by black-box metrics, no access to model internals required, and frontier models differ clearly in failure profile.

• Scaling splits "reading context" into two subskills moving in opposite directions. Google gives the first scaling law for contextual entrainment across two model families: the largest model resists counterfactual rumors 4x better than the smallest, but gets derailed by irrelevant tokens 2x more often.

• Pruning one objective leaves the better solution on the table: Google's MOONSHOT reframes post-training one-shot pruning as multi-objective optimization and wraps existing pruners, cutting C4 perplexity by up to 32.6% on Llama-3.2 at 2:4 sparsity.

Also Notable

• Microsoft Adds Video Grounding to Web Agent Skills — Text-only workflow descriptions leave too much execution ambiguity; visual demos anchor skills to UI elements. WebXSkill
• HETA Uses Hessian Second-Order Info to Fix Token Attribution — Existing methods mostly rely on linear approximations, missing the causal chain in autoregressive LLMs. Accepted to ICLR. HETA
• Procedurally Generated Open-Ended Science Reasoning Problems — Current benchmarks inherit biases from known paper conclusions; InfiniteScienceGym sidesteps publication bias and annotation noise. InfiniteScienceGym
• LLMs Fill Text Attributes for Medical Knowledge Graphs — Medical concept representations have long been limited by code noise and sparse samples; downstream clinical prediction quality improves substantially.
• MIT Builds a Mathematical Framework for t-SNE's Information Loss — Which structures are necessarily lost and which can be preserved, with a theoretical yardstick for the first time. Some Theoretical Limitations of t-SNE
• SSD-GS Fills In Material-Light Interaction for 3DGS Relighting — Physically accurate relighting requires modeling scattering and shadows separately. SSD-GS

Read the full edition →