From Context to Skills: Can Language Models Learn from Context Skillfully?

TL;DR

This paper introduces Ctx2Skill, a self-evolving framework enabling language models to autonomously discover and refine context-specific skills, enhancing their reasoning over complex, dense contexts without human supervision.

Contribution

It proposes a multi-agent self-play system with automated skill discovery and refinement, addressing manual annotation costs and lack of external feedback in context learning.

Findings

Improves solving rates on four CL-bench tasks across various models.

Automatically discovers and refines context-specific skills without human supervision.

Ensures robust skill evolution with a Cross-time Replay mechanism.

Abstract

Many real-world tasks require language models (LMs) to reason over complex contexts that exceed their parametric knowledge. This calls for context learning, where LMs directly learn relevant knowledge from the given context. An intuitive solution is inference-time skill augmentation: extracting the rules and procedures from context into natural-language skills. However, constructing such skills for context learning scenarios faces two challenges: the prohibitive cost of manual skill annotation for long, technically dense contexts, and the lack of external feedback for automated skill construction. In this paper, we propose Ctx2Skill, a self-evolving framework that autonomously discovers, refines, and selects context-specific skills without human supervision or external feedback. At its core, a multi-agent self-play loop has a Challenger that generates probing tasks and rubrics, a Reasoner that attempts to solve them guided by an evolving skill set, and a neutral Judge that provides binary feedback. Crucially, both the Challenger and the Reasoner evolve through accumulated skills: dedicated Proposer and Generator agents analyze failure cases and synthesize them into targeted skill updates for both sides, enabling automated skill discovery and refinement. To prevent adversarial collapse caused by increasingly extreme task generation and over-specialized skill accumulation, we further introduce a Cross-time Replay mechanism that identifies the skill set achieving the best balance across representative cases for the Reasoner side, ensuring robust and generalizable skill evolution. The resulting skills can be plugged into any language model to obtain better context learning capability. Evaluated on four context learning tasks from CL-bench, Ctx2Skill consistently improves solving rates across backbone models.