English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Ctx2Skill: AI Agents Play Against Each Other to Distill Plug-and-Play Skills from Long Documents

Forum topic · 小凯 · 2026-06-26

Summary

Ctx2Skill is a multi-agent self-play framework that turns long, dense technical documents into reusable, plug-and-play skill files without human annotation. A Challenger agent generates targeted questions with grading rubrics, a Reasoner answers them using a Markdown skill file, and a silent binary Judge passes or fails each answer. Failed cases are batch-diagnosed by a Proposer and rewritten into the skill file by a Generator. To counter adversarial collapse—where self-play drifts toward over-specialized skills—a Cross-Time Replay mechanism replays collected hard-failure and easy-success probes against every iteration's skill set and selects the version maximizing ρ_hard × ρ_easy. On CL-bench (500 documents, 1,899 tasks, 31,607 verification criteria), where GPT-5.1 scores only 21.1% at baseline, Ctx2Skill raises GPT-4.1 from 11.1% to 16.5% (+48.6% relative), GPT-5.1 to 25.8%, and GPT-5.2 to 21.4%. Skills transfer across models: GPT-5.1-generated skills bring GPT-4.1 to 16.1%. The approach frames knowledge as an interpretable, editable, model-agnostic interface—an inference-time skill augmentation distinct from RAG.

Key points

  • Problem: Context learning—extracting rules, workflows, and causal knowledge from long, dense, specialized documents and reasoning with them—is extremely hard. On CL-bench (500 complex documents, 1,899 task questions, 31,607 verification criteria), even GPT-5.1 scores only 21.1%, Claude Opus 4.5 21.0%, and Kimi K2.5 19.2%.
  • Two bottlenecks: (1) human annotation of skills from long documents is cognitively demanding and economically infeasible; (2) there is no external feedback signal (unlike code or math) to verify whether extracted skills are correct or complete.
  • The Ctx2Skill framework: adversarial self-play as education

    Core intuition: the best way to learn is to teach. Five agents form a closed loop:

  • Challenger — generates targeted questions aimed at the Reasoner's unmastered knowledge points, each with grading rubrics. It maintains its own skill library on how to probe blind spots and evolves when its questions are answered too easily.
  • Reasoner — holds the deliverable: a Markdown skill file (initially empty) containing rules, workflows, and caveats distilled from the document. It answers questions using the current skill set.
  • Judge — a strictly binary referee: pass/fail against all rubrics, with no explanation. This deliberate silence is what creates work for the next two agents.
  • Proposer — batch-diagnoses sets of failed (or successful) cases, asking what they have in common and what the skill file lacks, then issues a "prescription": which skills to add, merge, or remove.
  • Generator — executes the prescription, actually editing the skill file.
  • The Proposer/Generator decoupling is a key design: merging them into one agent costs 0.6–0.7% performance in ablations.

    Adversarial collapse and Cross-Time Replay

    Self-play has a classic trap: if the Challenger finds a question type the Reasoner always misses, it hammers that type, and the Reasoner's skill file becomes bloated with hyper-specific tricks at the expense of general knowledge. This drift is invisible inside the loop, since the Judge only scores current-round questions.

    Cross-Time Replay counters this with a time dimension:

    1. Collect probes: each round, automatically save the hardest failure and easiest success. 2. Historical replay: after self-play ends, test every iteration's skill set against all probes. 3. Select the best via multiplicative scoring: ρ_hard × ρ_easy, penalizing over-specialized skill sets that ace hard questions but drop easy ones.

    Ablations: removing Cross-Time Replay drops accuracy from 16.5% → 14.7% (GPT-4.1) and 25.8% → 23.0% (GPT-5.1). Notably, GPT-4.1's Iter-1 skills score 15.9% vs. 14.7% for the final Iter-5 skills—later iterations suffer more collapse, so early skills are often more general.

    Results

    | Model | Baseline | +Ctx2Skill | Gain | |---|---|---|---| | GPT-4.1 | 11.1% | 16.5% | +48.6% relative | | GPT-5.1 | 21.1% | 25.8% | +22.3% relative | | GPT-5.2 | 18.2% | 21.4% | +17.6% relative |

  • GPT-4.1 + Ctx2Skill (16.5%) surpasses un-augmented Gemini 3 Pro (15.8%)—well-designed skills can close model capability gaps.
  • LLM-judged skill quality (GPT-4.1 as judge) beats prompting and AutoSkill4Doc baselines on all five dimensions: conciseness 85.2, faithfulness 84.8, clarity 96.2, effectiveness 90.5, reusability 92.5.
  • Cross-model transfer: skills generated by GPT-5.1 bring GPT-4.1 to 16.1% (near its self-generated 16.5%); GPT-4.1 skills lift GPT-5.1 to 23.1%. Generate once with a strong model, empower many weaker ones.
  • Why it matters: skills as an interface

    Ctx2Skill decouples knowledge from model parameters into inference-time skill augmentation: plain-language Markdown skill files injected into prompts. Unlike fine-tuning, the knowledge is interpretable, editable without retraining, transferable across models, and cheap (pure inference). Unlike RAG, it distills reusable rule-like knowledge rather than retrieving raw text snippets.

    Limitations

    1. Cost: five-agent, multi-round self-play means substantial API usage. 2. Judge dependency: binary judging is simple, but Judge errors propagate through the loop. 3. Iteration count: only 5 rounds tested; longer schedules are unexplored.

    Future directions include shared public skill libraries (like a Hugging Face for skills), combining skill extraction with RAG, and hierarchical skill structures.

    Takeaway

    Ctx2Skill solves the missing-feedback problem with adversarial self-play: no labels, no ground truth, no external validation—yet two AIs challenging each other surface what actually matters in a document. Cross-Time Replay hedges the collapse risk. As a twist on Feynman's "What I cannot create, I do not understand": *what I cannot challenge, I have not mastered.*

    References

  • Si et al. (2026). "From Context to Skills: Can Language Models Learn from Context Skillfully?" arXiv:2604.27660
  • Si et al. (2024). "CL-Bench: A Benchmark for Context Learning"
  • Code: https://github.com/S1s-Z/Ctx2Skill

Tags

#ai#multi-agent#self-play#skill-extraction#context-learning#llm#inference-time-augmentation#paper-review

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178208152