IdeaCredIdeaCred

alphadl/AgentHER

71

AgentHER: Hindsight Experience Replay for LLM Agents

What's novel

AgentHER: Hindsight Experience Replay for LLM Agents

Code Analysis

16 files read · 4 rounds

A 4-stage pipeline that takes failed LLM agent trajectories, validates the failure, extracts actual achievements, reverse-engineers a new prompt the trajectory satisfies (with multi-judge verification), and packages the result into SFT/DPO/ShareGPT training formats.

Strengths

Clean separation of concerns with dual-mode (rule-based/LLM) operation at each stage, thoughtful multi-judge verification algorithm with temperature escalation and confidence merging, and robust structured-output parsing with multiple fallback strategies. The code is well-typed, well-documented, and the CLI is polished with rich output.

Weaknesses

No tests for the most complex stage (prompt_relabeler), and the rule-based failure detector's keyword matching is fairly naive (e.g., 'i believe' as a hallucination signal). The async client is defined but never used in the pipeline, suggesting incomplete integration.

Score Breakdown

Innovation
6 (25%)
Craft
73 (35%)
Traction
6 (15%)
Scope
67 (25%)

Signal breakdown

Innovation

Not Fork+1
Code Novelty+1
Concept Novelty+2

Craft

Ci+0
Tests+8
Polish+1
Releases+0
Has License+5
Code Quality+25
Readme Quality+15
Recent Activity+7
Structure Quality+5
Commit Consistency+2
Has Dependency Mgmt+5

Traction

Forks+0
Stars+6
Hn Points+0
Watchers+0
Early Traction+0
Devto Reactions+0
Community Contribs+0

Scope

Commits+7
Languages+3
Subsystems+10
Bloat Penalty+0
Completeness+7
Contributors+5
Authored Files+8
Readme Code Match+3
Architecture Depth+3
Implementation Depth+8

Evidence

Commits

44

Contributors

1

Files

29

Active weeks

3

TestsCI/CDREADMELicenseContributing

Repository

Language

Python

Stars

1

Forks

0

License

Apache-2.0