The Router Within: Eliciting Native Skill Routing from a Frozen LLM

• URL: http://arxiv.org/abs/2609.15982v1
• Date: Mon, 14 Sep 2026 17:58:27 GMT
• Title: The Router Within: Eliciting Native Skill Routing from a Frozen LLM
• Abstract summary: Skills extend an agent's parametric knowledge, and the gain they promise rests on picking the right one.<n>We show that the frozen agent LLM already carries the routing signal in its own forward passes.<n>A glance projects the task's and each skill's mid-layer states through the two maps, the only parameters trained.<n>A verdict then resumes the shortlisted skills' forward passes and reads the model's own likelihood and yes/no judgment.
• Score: 36.31894794808421
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Skills extend an LLM agent beyond its parametric knowledge, and the gain they promise rests on picking the right one. Deployed harnesses route by preloading every skill's metadata into the context, which disperses the agent's attention and caps the library size. Retrieval pipelines move the selection out of the context, but also out of the agent's capability. We show that the frozen agent LLM already carries the routing signal in its own forward passes, and that two linear maps suffice to read it out with no skill text in the context. Gavel (Glance And Verdict from a frozen LLM) reads it in two steps. A glance projects the task's and each skill's mid-layer states through the two maps, the only parameters trained, and scores the full library against compact per-skill banks that one forward pass builds at installation. A verdict then resumes the shortlisted skills' forward passes and reads the model's own likelihood and yes/no judgment, fused with the glance as a product of experts. Trained once, Gavel transfers zero-shot to three public benchmarks and SkillTraj, our new benchmark of 372 simulated agent trajectories. On Qwen3-32B it outperforms progressive disclosure and retrieve-and-rerank pipelines that add 1.2B to 16B external parameters, by up to 13.4 points on written tasks and up to 21.9 when the need for a skill arises mid-rollout. Routing accuracy improves as the backbone does, and in a bash-agent harness the same 32B triggers the correct skill on Skill-Use more often than far larger frontier models running in Codex.

Related papers

• Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills [86.15447717549154]
  We present DisCo, a skill-powered research agent that creates skills and uses them during research.<n>We call this missing layer operational knowledge, the know-how that separates knowing a method from making it work.
  arXiv  Detail & Related papers  (2026-09-02T15:49:41Z)
• Anything2Skill: Compiling External Knowledge into Reusable Skills for Agents [27.965275812765015]
  Anything2Skill is a taxonomy-guided framework that compiles heterogeneous external knowledge into reusable, retrievable, and executable skills for agents.<n>We show that Anything2Skill combined with RAG achieves 98.85% and 94.10% success rates, respectively.
  arXiv  Detail & Related papers  (2026-06-08T10:24:32Z)
• Skill Is Not Document: A Query-Conditional Benchmark and Two-Stage Retriever for LLM Agent Skill Routing [40.648572239231804]
  R3-Skill is a benchmark for realistic agent skill routing.<n>We build a two-stage retrieval system with skill compatibility as an explicit training signal.<n>The dataset, training code and model weights are released as open source for agent skill routing.
  arXiv  Detail & Related papers  (2026-06-02T12:30:46Z)
• COLLEAGUE.SKILL: Automated AI Skill Generation via Expert Knowledge Distillation [58.84646485020439]
  We present an automated trace-to-skill distillation system for generating person-grounded AI skills.<n> COLLEAGUE.SKILL produces a versioned skill package with two tracks: a capability track for practices, mental models, and decisions, and a bounded behavior track for communication style, interaction rules, and correction history.
  arXiv  Detail & Related papers  (2026-05-29T12:59:08Z)
• Skill-as-Pseudocode: Refactoring Skill Libraries to Pseudocode for LLM Agents [50.78646963535282]
  We propose Skill-as-Pseudocode (SaP), an automatic conversion of markdown skill libraries into typed pseudocode.<n>For each cluster of similar procedural passages drawn from one or more skills, SaP extracts a typed contract and filters it through a four-check deterministic verifier.<n>Promoted contracts are inlined into a rewritten skill skeleton together with restored concrete action templates, giving the agent two complementary signals.
  arXiv  Detail & Related papers  (2026-05-27T04:48:40Z)
• Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills [21.971456179996093]
  We introduce Trace2Skill, a framework that mirrors how human experts author skills.<n>Instead of reacting sequentially to individual trajectories, Trace2Skill dispatches a parallel fleet of sub-agents to analyze a diverse pool of executions.<n>It extracts trajectory-specific lessons and hierarchically consolidates them into a unified, conflict-free skill directory via inductive reasoning.
  arXiv  Detail & Related papers  (2026-03-26T08:26:38Z)
• SkillRouter: Skill Routing for LLM Agents at Scale [18.540967600258607]
  Given a user task, the system must identify relevant skills before downstream planning or execution.<n>Existing agent stacks often rely on progressive disclosure, exposing only skill names and descriptions while hiding the full implementation body.<n>We present Skill, a compact 1.2B full-text retrieve-and-rerank pipeline.
  arXiv  Detail & Related papers  (2026-03-23T18:23:59Z)
• A Tale of LLMs and Induced Small Proxies: Scalable Agents for Knowledge Mining [52.57170634957137]
  Falconer is a framework that combines large language models with lightweight proxy models for scalable knowledge mining.<n>LLMs act as planners, decomposing user instructions into executable pipelines, and as annotators, generating supervision to train small proxies.<n>Experiments show that Falconer closely matches state-of-the-art LLMs in instruction-following accuracy while reducing inference cost by up to 90%.
  arXiv  Detail & Related papers  (2025-10-01T20:06:48Z)
• Get my drift? Catching LLM Task Drift with Activation Deltas [55.75645403965326]
  Task drift allows attackers to exfiltrate data or influence the LLM's output for other users.<n>We show that a simple linear classifier can detect drift with near-perfect ROC AUC on an out-of-distribution test set.<n>We observe that this approach generalizes surprisingly well to unseen task domains, such as prompt injections, jailbreaks, and malicious instructions.
  arXiv  Detail & Related papers  (2024-06-02T16:53:21Z)
• Self-Prompting Large Language Models for Zero-Shot Open-Domain QA [67.08732962244301]
  Open-Domain Question Answering (ODQA) aims to answer questions without explicitly providing background documents. This task becomes notably challenging in a zero-shot setting where no data is available to train tailored retrieval-reader models. We propose a Self-Prompting framework to explicitly utilize the massive knowledge encoded in the parameters of Large Language Models.
  arXiv  Detail & Related papers  (2022-12-16T18:23:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.