SoftSkill: Behavioral Compression for Contextual Adaptation

2026-06-18

AI Summary

This work addresses the inefficiency of conventional natural language–based Markdown agents, whose skills must be repeatedly parsed by a frozen language model for each task, leading to suboptimal behavioral guidance. To overcome this limitation, the authors propose SoftSkill, a novel approach that compresses natural language skills into short sequences of soft prompts—comprising only 32 virtual tokens—which serve as trainable continuous context vectors. These soft prompts are optimized via next-token prediction within a frozen large language model, thereby encoding lightweight yet effective implicit behavioral priors. Evaluated on Qwen1.5-4B, SoftSkill substantially outperforms both no-skill prompting and the SkillOpt baseline, achieving performance gains of 8.3, 42.1, and 1.3 points on SearchQA, LiveMath, and DocVQA benchmarks, respectively.

Abstract

Agent skills are commonly deployed as natural-language Markdown files that encode answer policies, evidence-use habits, and task procedures. These files are readable and portable, but they are consumed indirectly: for each task instance, a frozen language model must translate a long textual artifact into generation-time behavior. This paper asks whether a natural-language skill can instead initialize a compact continuous context object, refined by a trainable soft delta while the base model remains frozen. We propose SoftSkill, a frozen-backbone method that tunes such soft skills with next-token prediction and deploys them as latent behavioral priors at inference time. In our main single-round setting, a length-32 SoftSkill prefix on Qwen3.5-4B improves over no-skill prompting by 8.3 points on SearchQA, 42.1 points on LiveMath, and 1.3 points on DocVQA. Relative to SkillOpt, SoftSkill improves accuracy by 5.2 points on SearchQA and 12.5 points on LiveMath, while replacing hundreds to thousands of Markdown skill tokens with a few virtual tokens. We further study agentic execution as a harder boundary case, where sparse trajectory imitation provides useful signal but does not yet robustly compress long-horizon procedural behavior. More broadly, the results suggest that some task skills are better treated not as additional Markdown to be reinterpreted at inference time, but as compact latent controls over how a frozen model enters the task.

Problem

• behavioral compression
• contextual adaptation
• soft skills
• frozen language models
• skill representation

Innovation

• behavioral compression
• soft prompting
• frozen language models
• latent skill representation
• contextual adaptation