SkillComposer performs task-conditioned skill sequence prediction with a constrained autoregressive decoder to jointly output skill subset, count, and order, raising pass rates by 23.1 and 18.2 percentage points on two production coding agents over no-skill baselines.
hub Canonical reference
SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning
Canonical reference. 82% of citing Pith papers cite this work as background.
58
Pith papers citing it
Background
82% of classified citations
abstract
Large Language Model (LLM) agents have shown stunning results in complex tasks, yet they often operate in isolation, failing to learn from past experiences. Existing memory-based methods primarily store raw trajectories, which are often redundant and noise-heavy. This prevents agents from extracting high-level, reusable behavioral patterns that are essential for generalization. In this paper, we propose SkillRL, a framework that bridges the gap between raw experience and policy improvement through automatic skill discovery and recursive evolution. Our approach introduces an experience-based distillation mechanism to build a hierarchical skill library SkillBank, an adaptive retrieval strategy for general and task-specific heuristics, and a recursive evolution mechanism that allows the skill library to co-evolve with the agent's policy during reinforcement learning. These innovations significantly reduce the token footprint while enhancing reasoning utility. Experimental results on ALFWorld, WebShop and seven search-augmented tasks demonstrate that SkillRL achieves state-of-the-art performance, outperforming strong baselines over 15.3% and maintaining robustness as task complexity increases. Code is available at this https://github.com/aiming-lab/SkillRL.
hub tools
citation-role summary
background 14
baseline 1
dataset 1
other 1
citation-polarity summary
years
2026 58representative citing papers
SkillEvolBench is a new diagnostic benchmark that evaluates the transition from episodic experience to procedural skills in LLM agents using role-conditioned task families and frozen deployment tests.
Residual skill optimization creates complementary Text-to-SQL agents by training each new skill on prior ensemble failures, yielding accuracy gains on Spider2-Lite and transfer to other dialects and tasks.
Terminal-World is a skill-based synthesis pipeline that generates 5,723 training environments and produces Terminal-World-32B which outperforms baselines on Terminal-Bench 2.0 using only 1.2% of the data.
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
AEvo introduces a meta-agent that edits the evolution procedure or agent context based on accumulated state, outperforming baselines by 26% relative improvement on agentic benchmarks and achieving SOTA on open-ended tasks.
EvolveMem enables autonomous self-evolution of LLM memory retrieval configurations via LLM diagnosis and safeguards, delivering 25.7% gains over strong baselines on LoCoMo and 18.9% on MemBench with positive cross-benchmark transfer.
MMSkills packages multimodal procedural knowledge into state-conditioned skills with text, state cards, and multi-view keyframes, generated from public trajectories via an agentic process and used at inference via branch-loaded inspection to improve visual agents on GUI and game benchmarks.
OLIVIA treats LLM agent action selection as a contextual linear bandit over frozen hidden states and applies UCB exploration to adapt online, yielding consistent gains over static ReAct and prompt-based baselines on four benchmarks.
Evolving-RL jointly optimizes experience extraction and utilization in LLM agents via RL with separate evaluation signals, delivering up to 98.7% relative gains on out-of-distribution tasks in ALFWorld and Mind2Web.
RewardHarness self-evolves a tool-and-skill library from 100 preference examples to reach 47.4% accuracy on image-edit evaluation, beating GPT-5, and yields stronger RL-tuned models.
AHE automates coding-agent harness evolution via component, experience, and decision observability, raising Terminal-Bench 2 pass@1 from 69.7% to 77.0% with cross-benchmark and cross-model transfer.
COSPLAY co-evolves an LLM decision agent with a skill bank agent to improve long-horizon game performance, reporting over 25.1% average reward gains versus frontier LLM baselines on single-player benchmarks.
GraSP introduces executable skill graphs that improve LLM agent rewards by up to 19 points and reduce steps by up to 41% over ReAct, Reflexion, ExpeL, and flat-skill baselines across ALFWorld, ScienceWorld, WebShop, and InterCode.
SkillFlow benchmark shows lifelong skill evolution yields modest gains for some models like Claude Opus 4.6 but limited or negative utility for others despite high skill usage.
SAGER equips LLM recommendation agents with per-user evolving policy skills via two-representation architecture, contrastive CoT diagnosis, and skill-augmented listwise reasoning, yielding SOTA gains orthogonal to memory accumulation.
This survey introduces the Generate-Filter-Control-Replay (GFCR) taxonomy to structure rollout pipelines for RL-based post-training of reasoning LLMs.
SkillRevise iteratively refines initial LLM-generated agent skills using execution traces to diagnose defects and apply repairs, raising success rates from 36.05% to 61.63% on SkillsBench across three benchmarks and five LLMs.
SkillBrew introduces a Pareto-aware multi-objective optimization framework with bi-level propose-then-verify to curate skill banks for LLM agents, evaluated on two public benchmarks.
SGSD retrieves skill-mistake pairs to build a multi-teacher pool, validates teacher polarity via a verifier, and applies a gated objective to distill useful signals, yielding 6.2% average gains over GRPO on math benchmarks with Qwen3-1.7B.
SkillC converts skill-helpfulness contrast into a policy learning signal via paired rollouts and dual-stream advantage estimation, outperforming prior internalization baselines by 5.5% and 4.4% on ALFWorld and WebShop without runtime skill access.
ClueAegis introduces a heuristic-to-reasoning cognitive skill framework and ClueAegis-Bench for evidence-based synthetic image detection that outperforms end-to-end classifiers in generalization and explainability.
DemoEvolve bootstraps harness evolution with demonstrations to achieve more stable and effective edits than self-rollout search in sparse-feedback environments like Balatro.
SEAL co-evolves LLM agents and environments via shared turn-level failure diagnoses, yielding +8.25 to +26.25 point gains on tool-use tasks with only 400 samples.
citing papers explorer
-
Generative Skill Composition for LLM Agents
SkillComposer performs task-conditioned skill sequence prediction with a constrained autoregressive decoder to jointly output skill subset, count, and order, raising pass rates by 23.1 and 18.2 percentage points on two production coding agents over no-skill baselines.
-
SkillEvolBench: Benchmarking the Evolution from Episodic Experience to Procedural Skills
SkillEvolBench is a new diagnostic benchmark that evaluates the transition from episodic experience to procedural skills in LLM agents using role-conditioned task families and frozen deployment tests.
-
Residual Skill Optimization for Text-to-SQL Ensembles
Residual skill optimization creates complementary Text-to-SQL agents by training each new skill on prior ensemble failures, yielding accuracy gains on Spider2-Lite and transfer to other dialects and tasks.
-
Terminal-World: Scaling Terminal-Agent Environments via Agent Skills
Terminal-World is a skill-based synthesis pipeline that generates 5,723 training environments and produces Terminal-World-32B which outperforms baselines on Terminal-Bench 2.0 using only 1.2% of the data.
-
ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
-
Harnessing Agentic Evolution
AEvo introduces a meta-agent that edits the evolution procedure or agent context based on accumulated state, outperforming baselines by 26% relative improvement on agentic benchmarks and achieving SOTA on open-ended tasks.
-
EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents
EvolveMem enables autonomous self-evolution of LLM memory retrieval configurations via LLM diagnosis and safeguards, delivering 25.7% gains over strong baselines on LoCoMo and 18.9% on MemBench with positive cross-benchmark transfer.
-
MMSkills: Towards Multimodal Skills for General Visual Agents
MMSkills packages multimodal procedural knowledge into state-conditioned skills with text, state cards, and multi-view keyframes, generated from public trajectories via an agentic process and used at inference via branch-loaded inspection to improve visual agents on GUI and game benchmarks.
-
OLIVIA: Online Learning via Inference-time Action Adaptation for Decision Making in LLM ReAct Agents
OLIVIA treats LLM agent action selection as a contextual linear bandit over frozen hidden states and applies UCB exploration to adapt online, yielding consistent gains over static ReAct and prompt-based baselines on four benchmarks.
-
Evolving-RL: End-to-End Optimization of Experience-Driven Self-Evolving Capability within Agents
Evolving-RL jointly optimizes experience extraction and utilization in LLM agents via RL with separate evaluation signals, delivering up to 98.7% relative gains on out-of-distribution tasks in ALFWorld and Mind2Web.
-
RewardHarness: Self-Evolving Agentic Post-Training
RewardHarness self-evolves a tool-and-skill library from 100 preference examples to reach 47.4% accuracy on image-edit evaluation, beating GPT-5, and yields stronger RL-tuned models.
-
Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
AHE automates coding-agent harness evolution via component, experience, and decision observability, raising Terminal-Bench 2 pass@1 from 69.7% to 77.0% with cross-benchmark and cross-model transfer.
-
Co-Evolving LLM Decision and Skill Bank Agents for Long-Horizon Tasks
COSPLAY co-evolves an LLM decision agent with a skill bank agent to improve long-horizon game performance, reporting over 25.1% average reward gains versus frontier LLM baselines on single-player benchmarks.
-
GraSP: Graph-Structured Skill Compositions for LLM Agents
GraSP introduces executable skill graphs that improve LLM agent rewards by up to 19 points and reduce steps by up to 41% over ReAct, Reflexion, ExpeL, and flat-skill baselines across ALFWorld, ScienceWorld, WebShop, and InterCode.
-
SkillFlow:Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agents
SkillFlow benchmark shows lifelong skill evolution yields modest gains for some models like Claude Opus 4.6 but limited or negative utility for others despite high skill usage.
-
SAGER: Self-Evolving User Policy Skills for Recommendation Agent
SAGER equips LLM recommendation agents with per-user evolving policy skills via two-representation architecture, contrastive CoT diagnosis, and skill-augmented listwise reasoning, yielding SOTA gains orthogonal to memory accumulation.
-
Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning
This survey introduces the Generate-Filter-Control-Replay (GFCR) taxonomy to structure rollout pipelines for RL-based post-training of reasoning LLMs.
-
SkillRevise: Improving LLM-Authored Agent Skills via Trace-Conditioned Skill Revision
SkillRevise iteratively refines initial LLM-generated agent skills using execution traces to diagnose defects and apply repairs, raising success rates from 36.05% to 61.63% on SkillsBench across three benchmarks and five LLMs.
-
SkillBrew: Multi-Objective Curation of Skill Banks for LLM Agents
SkillBrew introduces a Pareto-aware multi-objective optimization framework with bi-level propose-then-verify to curate skill banks for LLM agents, evaluated on two public benchmarks.
-
Skill-Conditioned Gated Self-Distillation for LLM Reasoning
SGSD retrieves skill-mistake pairs to build a multi-teacher pool, validates teacher polarity via a verifier, and applies a gated objective to distill useful signals, yielding 6.2% average gains over GRPO on math benchmarks with Qwen3-1.7B.
-
SKILLC: Learning Autonomous Skill Internalization in LLM Agents via Contrastive Credit Assignment
SkillC converts skill-helpfulness contrast into a policy learning signal via paired rollouts and dual-stream advantage estimation, outperforming prior internalization baselines by 5.5% and 4.4% on ALFWorld and WebShop without runtime skill access.
-
ClueAegis: Heuristic-to-Reasoning Cognitive-skill Learning for Unified Evidence-based Synthetic Image Detection
ClueAegis introduces a heuristic-to-reasoning cognitive skill framework and ClueAegis-Bench for evidence-based synthetic image detection that outperforms end-to-end classifiers in generalization and explainability.
-
DemoEvolve: Overcoming Sparse Feedback in Agentic Harness Evolution with Demonstrations
DemoEvolve bootstraps harness evolution with demonstrations to achieve more stable and effective edits than self-rollout search in sparse-feedback environments like Balatro.
-
SEAL: Synergistic Co-Evolution of Agents and Learning Environments
SEAL co-evolves LLM agents and environments via shared turn-level failure diagnoses, yielding +8.25 to +26.25 point gains on tool-use tasks with only 400 samples.
-
SkillOpt: Executive Strategy for Self-Evolving Agent Skills
SkillOpt introduces a controllable text-space optimizer that evolves agent skills via add/delete/replace edits accepted only on strict held-out validation improvement, reporting consistent gains across 52 model-benchmark-harness combinations.
-
From Raw Experience to Skill Consumption: A Systematic Study of Model-Generated Agent Skills
A systematic study across five domains finds model-generated skills yield average gains but non-uniform negative transfer, with a meta-skill improving extraction quality.
-
Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles
Maestro uses outcome-based RL to train a lightweight policy that orchestrates ensembles of frozen expert models and skills, reporting 70.1% average accuracy across ten multimodal benchmarks and outperforming GPT-5 and Gemini-2.5-Pro while generalizing to unseen components.
-
Ratchet: A Minimal Hygiene Recipe for Self-Evolving LLM Agents
Ratchet provides a minimal hygiene recipe for self-managing skill libraries in frozen LLM agents, delivering +0.328 rolling-mean pass@1 gain on MBPP+ hard-100 and +0.22 peak lift on SWE-bench Verified.
-
Is One Score Enough? Rethinking the Evaluation of Sequentially Evolving LLM Memory
SeqMem-Eval reveals that high final accuracy in sequential LLM memory tasks often coexists with substantial forgetting and negative transfer, exposing stability-adaptability trade-offs hidden by standard aggregate metrics.
-
MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning
MAP improves LLM agent reasoning by constructing a structured cognitive map of the environment before task execution, yielding performance gains on benchmarks like ARC-AGI-3 and superior training data via the new MAP-2K dataset.
-
SkillGraph: Skill-Augmented Reinforcement Learning for Agents via Evolving Skill Graphs
SkillGraph represents skills as nodes in an evolving directed graph with typed dependency edges and updates the graph from RL trajectories to boost compositional task performance.
-
SkillEvolver: Skill Learning as a Meta-Skill
A meta-skill authors and refines prose-and-code skills for agents by learning from post-deployment failures with an overfit audit, achieving 56.8% accuracy on SkillsBench tasks versus 43.6% for human-curated skills.
-
HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
HAGE proposes a trainable weighted graph memory framework with LLM intent classification, dynamic edge modulation, and RL optimization that improves long-horizon reasoning accuracy in agentic LLMs over static baselines.
-
Evidence Over Plans: Online Trajectory Verification for Skill Distillation
SPARK generates environment-verified trajectories to compute PDI, enabling posterior skill distillation that outperforms no-skill baselines and human-written skills across 86 tasks with up to 1000x cheaper inference.
-
ORACLE: Anticipating Scams from Partial Trajectories in Streaming App Usage
ORACLE is a new agentic framework using adaptive context consolidation and teacher-student distillation to detect emerging scam patterns from incomplete, long-horizon app usage streams across 12 scam types.
-
SearchSkill: Teaching LLMs to Use Search Tools with Evolving Skill Banks
SearchSkill improves LLM query planning on knowledge QA by using explicit skill selection from an evolving SkillBank and a two-stage SFT process that aligns training with inference-time skill-grounded execution.
-
SkillMaster: Toward Autonomous Skill Mastery in LLM Agents
SkillMaster enables LLM agents to autonomously develop skills via trajectory review, counterfactual evaluation, and DualAdv-GRPO training, boosting success rates by 8.8% on ALFWorld and 9.3% on WebShop.
-
SkillOS: Learning Skill Curation for Self-Evolving Agents
SkillOS is an RL recipe that learns to curate reusable skills for self-evolving LLM agents, outperforming memory-free and memory-based baselines while generalizing across executors and domains.
-
From Context to Skills: Can Language Models Learn from Context Skillfully?
Ctx2Skill uses a self-evolving multi-agent loop with Challenger, Reasoner, Judge, and Cross-time Replay to discover context-specific skills, improving task-solving rates on CL-bench benchmarks across models.
-
ClawTrace: Cost-Aware Tracing for LLM Agent Skill Distillation
ClawTrace enables cost-aware LLM agent skill distillation by tracing per-step costs and generating preserve, prune, and repair patches, with ablations showing reduced regressions and prune rules transferring to cut costs by 32%.
-
SkillGraph: Self-Evolving Multi-Agent Collaboration with Multimodal Graph Topology
SkillGraph jointly evolves agent skills and collaboration topologies in multi-agent vision-language systems using a multimodal graph transformer and a skill designer, yielding consistent performance gains on benchmarks.
-
GrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning
GrandCode is the first AI system to consistently beat all human participants and place first in live Codeforces competitive programming contests.
-
The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory
Janus is a method-agnostic plug-in that uses a Memory Momentum Trigger and compact hybrid evaluation to selectively accept LLM memory updates, yielding +2.7 to +4.6 accuracy gains over base updaters on six datasets.
-
Parametric Skills
ParametricSkills uses a hypernetwork to turn textual skills into LoRA adapters, outperforming in-context learning by 6.44 points on average across six SWE subtasks with higher BERT Score and F1.
-
UCOB: Learning to Utilize and Evolve Agentic Skills via Credit-Aware On-Policy Bidirectional Self-Distillation
UCOB improves agentic RL by using return-to-go comparisons between skill-conditioned and no-skill prompts as local teachers for bidirectional self-distillation and skill memory updates.
-
SkillSmith: Co-Evolving Skills and Tools for Self-Improving Agent Systems
SkillSmith introduces a synergy-aware skill-tool co-evolution framework with atomic bundles, Lotka-Volterra-inspired interaction modeling, and anti-pattern recording that outperforms baselines on complex tasks.
-
Learning Design Skills as Memory Policies for Agentic Photonic Inverse Design
SkillPCF is a closed-loop agent framework with a physics-guided memory skill bank, reinforcement-learned skill selection, and simulator-grounded evolution that improves design quality and efficiency for photonic crystal fiber inverse design under limited simulation budgets.
-
AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration
AutoResearchClaw introduces a multi-agent research pipeline with debate, self-healing, verifiable outputs, human collaboration modes, and cross-run evolution that outperforms AI Scientist v2 by 54.7% on ARC-Bench.
-
MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization
MOCHA combines Chebyshev scalarization with exponential annealing to optimize LLM agent skills across performance and platform constraints, improving mean correctness by 7.5% over baselines on six tasks while finding more Pareto-optimal variants.
-
AtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM Agents
AtlasVA organizes VLM agent memory into spatial heatmaps, visual exemplars, and symbolic skills, evolving atlases from trajectories to act as potential-based shaping rewards in teacher-free reinforcement learning.