• Task Success Does Not Prove Good Multi-Agent Collaboration: AgentWorld tests 3–20 agents with asymmetric roles across 100 human-annotated tasks. The best model succeeds only 52.0% of the time. Its CCE metric traces causal dependencies between actions and measures how much team effort contributes to the outcome.

• PISA Cuts Sparse Block Selection to (O(N\log N)): Hierarchical Top-K selection and fused Triton kernels avoid materializing the full score matrix. The abstract provides no end-to-end latency or memory results, so the engineering gains still need validation.

• Natural-Image Priors Can Improve Elevation Maps: Researchers prune Stable Diffusion 3's text stream and condition it on photogrammetric DSMs and Pléiades imagery. In held-out Bordeaux, Dense Urban RMSE falls from 4.16 meters to 2.77 meters.

Also Notable

• Chess, Poker, and Werewolf Test Strategy Across Extended Games — Game Arena covers perfect-information, imperfect-information, and multiplayer games. It tests planning, adaptation, and reliability under uncertainty.
• Equivalent Softmax Reparameterization Reduces W4 Quantization Error — A separate result finds that quantizing Phi's output head cuts batch-one generation latency by 10.8% against a BF16 output-head baseline.
• Externalizing CPDAG State Improves Answers to Graph Queries — On Corr2Cause's primary paired evaluation, typed and schema-constrained CPDAG summaries raise Qwen3.5-27B's (F_1)(Yes) from 73.0 to 86.4.
• Misaligned Audio-Video Links Drive Three-Modal Binding Failures — Adding existing detection boxes around active speakers improves results across four conversational-video benchmarks without training.
• Traceable Evidence and Query-Aware Summaries Cut Output Budgets — H2S-14B retains 97.1% of its 16K-budget performance with a 4K output budget.
• Compressing Only Older Tool Observations Creates a Tunable Tradeoff — LOHA keeps the agent's actions and latest (K) observations verbatim. On SWE-bench Verified, larger (K) values generally trade less compression for higher solve rates.
• SatNav Builds 118,000 Drone Tasks Across 18 Cities — Three task types test loop-progress tracking, landmark-based localization, and route following with counting cues. The average trajectory spans 379 meters.
• Point Clouds and Gaussian Splats Become JSON Scene Graphs — GraphWrit3R avoids multi-stage pipelines that depend on ground-truth object annotations. It accepts point clouds, Gaussian Splats, or both.
• Hotel Price Sensitivity Varies More Than Tenfold Across Models — Across 28 models, average booked room prices range from $247 to $393. Measure each purchasing agent directly; capability scores and model families do not predict its preferences.
• Low-Dimensional State Enables One-Step Bayesian Online Adaptation — AURA learns a latent state space offline. It then uses an extended Kalman filter to update the state and reconstruct full parameters in non-stationary environments.

Read the full edition →