benchmark compilation 1capability probing 1evaluation methods 1large language models 1representation learning 1
From the 1 of 3 linked papers with an AI index.
3 papers
cs.CL2026
RepBench: Compiling Benchmarks into Capability Representations for Large Language Models
Yanshi Li, Xueru Bai, Shuman Liu +1
The paper introduces RepBench, a framework that aggregates thousands of benchmark datasets into a large set of probe texts to evaluate capability-aligned representations of large l…
cs.CL2026
Decomposing and Steering Functional Metacognition in Large Language Models
Yanshi Li, Xueru Bai, Shuman Liu +2
Large language models (LLMs) increasingly exhibit behaviors suggesting awareness of their evaluation context, often adapting their reasoning strategies in benchmark settings. Prior…
cs.AI2025
Each Prompt Matters: Scaling Reinforcement Learning Without Wasting Rollouts on Hundred-Billion-Scale MoE
Anxiang Zeng, Haibo Zhang, Hailing Zhang +13
We present CompassMax-V3-Thinking, a hundred-billion-scale MoE reasoning model trained with a new RL framework built on one principle: each prompt must matter. Scaling RL to this s…