Entity · paper

The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agents

paperactiveprovisionalthe-regression-tax-decomposing-why-skills-help-and-hurt-llm-agents-d5ac80b9·1 events·first seen 18h ago

Aliases: The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agents

More like this (12)

Generative Skill Composition for LLM Agents Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills Progress Advantage for LLM Agents Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill Multi-Component LLM Agent Model-Generated Agent Skills (paper)Role-Agent: Bootstrapping LLM Agents via Dual-Role Evolution Notes to Self: Can LLMs Benefit from Experiential Abstractions?LLM Bargaining Agents Teaching LLMs to Self-Evolve: Cultivating Core Meta-Skills with Reinforcement Learning LLM Agent Classroom

Recent events (1)

6arXiv · cs.AI·18h ago·source ↗

The Regression Tax: Why Adding Skills to LLM Agents Can Make Them Worse

A new arXiv paper introduces the concept of the 'regression tax' — the phenomenon where adding procedural skills to LLM agents causes previously-solved tasks to fail, even when the skill is never invoked. Across nearly 6,000 runs on two office automation benchmarks and three model harness stacks, the authors identify three regression mechanisms: skill description osmosis (context pollution), grounding displacement (procedure overrides input interpretation), and verification displacement (procedure suppresses output checking). The key finding is that top-performing skill configurations succeed primarily by regressing less, not by gaining more, and that grounding and verification are more important to reliability than procedural skill choice.

Evaluation and Benchmarking Agent and Tool Ecosystem The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agents