Feed

675 materials

Sort by
News19 November 2026

The Evolution of Agent Benchmarks as Task Pools Scale: The Role of Evidence Selection and Sequential Races

This work explains why increasing the number of tasks can hinder the evolution of agent benchmarks: the problem is poor evidence selection, not data volume itself. The proposed Successive Scale Racing iteratively selects diverse task combinations and compares candidates at low cost; across three benchmarks, it increased the share of successful runs as the pool grew from 40 to 200, while baseline methods degraded, and reduced validation runs by about 60%.

The Evolution of Agent Benchmarks as Task Pools Scale: The Role of Evidence Selection and Sequential Races
Read
News19 November 2026

Constraints in a Differentiable Graph Solver for Modeling and Inverse Problems

An implicit graph solver is introduced that supports end-to-end differentiation and accounts for constraints represented by a large system of equations. It has been tested on forward transient and steady-state simulations, coefficient and right-hand-side recovery, as well as learnable graph coarsening; the results indicate lower forward-simulation costs and more efficient execution of inverse problems.

Constraints in a Differentiable Graph Solver for Modeling and Inverse Problems
Read
News18 November 2026

Multimodal Multi-Agent LLMs in Mathematics: How to Verify Reasoning Steps and Robustness to Distortions

MAMath-Bench is introduced as a benchmark of 1,003 problems across 11 areas of mathematics, featuring text, graphics, and audio; in addition to the final answer, it records intermediate steps and the system’s response to adversarial changes in the reasoning process. The results show that collaboration among multiple LLMs does not necessarily improve accuracy, may amplify redundant justifications, and remains vulnerable to subtle structural perturbations.

Multimodal Multi-Agent LLMs in Mathematics: How to Verify Reasoning Steps and Robustness to Distortions
Read
News18 November 2026

Verifiable Tasks and Fine-Tuning on Long Trajectories: How Navalia Strengthens Cooperative LLM Agents

The paper describes a framework that uses corporate financial reports to build computational chains with traceable evidence, independently validates them, and then applies them for SFT and RL. Based on Qwen3.6-35B-A3B, this pipeline improves performance on five benchmarks and, according to the authors, enhances evidence linking, tool use, and long-horizon task execution, including office scenarios.

Verifiable Tasks and Fine-Tuning on Long Trajectories: How Navalia Strengthens Cooperative LLM Agents
Read