---
format: "aidr-story-markdown/v1"
id: "fce31a08cbec008afc908ae1857e6f2b37afdf646e858519a66a7eec5d748459"
canonical_url: "https://aidr.today/fce31a08?lang=vi"
title: "The Resolution Hypothesis: Agent Evolution's Key Is Identifying Improvements"
lang: "en"
requested_lang: "vi"
available_langs: ["en"]
translation_fallback: "en"
fallback_fields: ["title","summary"]
published_at: "2026-11-19T00:00:00.000Z"
category: "Research"
topics: ["agent"]
source_urls: ["https://huggingface.co/papers/2611.23068"]
summary: "Agent evolution follows a simple loop: a proposer writes candidate artifacts (prompts, skills, memories) and selection accepts those that score higher. This loop has shown promise, yet practitioners face a puzzle: methods trade wins across papers, and evolution often fails to bootstrap on certain tasks. We trace this puzzle to the Resolution Hypothesis: agent evolution stalls not because methods fail to propose better artifacts, but because noisy evaluations cannot reliably identify them. Across seven evolution methods and six models (over 10M LLM calls), we find support for the hypothesis at both levels where evaluation scores are used. At the candidate level (choosing artifacts), helpful candidate artifacts exist, but selection is no better than random at telling them from harmful ones, so errors accumulate across rounds. At the method level (ranking methods), typical margins of a few percentage points are smaller than the difference between two independent runs of the same method, so published rankings often rest on luck. Together, these findings indicate that agent evolution is often bottlenecked by reliably identifying improvements."
---

# The Resolution Hypothesis: Agent Evolution's Key Is Identifying Improvements

> [Open the canonical story](<https://aidr.today/fce31a08?lang=vi>)

**Published:** 2026-11-19T00:00:00.000Z
**Category:** Research
**Topics:** agent

_English fallback used for: title, summary._

## Summary

Agent evolution follows a simple loop: a proposer writes candidate artifacts \(prompts, skills, memories\) and selection accepts those that score higher\. This loop has shown promise, yet practitioners face a puzzle: methods trade wins across papers, and evolution often fails to bootstrap on certain tasks\. We trace this puzzle to the Resolution Hypothesis: agent evolution stalls not because methods fail to propose better artifacts, but because noisy evaluations cannot reliably identify them\. Across seven evolution methods and six models \(over 10M LLM calls\), we find support for the hypothesis at both levels where evaluation scores are used\. At the candidate level \(choosing artifacts\), helpful candidate artifacts exist, but selection is no better than random at telling them from harmful ones, so errors accumulate across rounds\. At the method level \(ranking methods\), typical margins of a few percentage points are smaller than the difference between two independent runs of the same method, so published rankings often rest on luck\. Together, these findings indicate that agent evolution is often bottlenecked by reliably identifying improvements\.

## Sources

- [Story source](<https://huggingface.co/papers/2611.23068>)

