Add ProposeCurateEngine: unified evolution engine for OSWorld & CL-bench#20
Open
weitianxin wants to merge 2 commits intomainfrom
Open
Add ProposeCurateEngine: unified evolution engine for OSWorld & CL-bench#20weitianxin wants to merge 2 commits intomainfrom
weitianxin wants to merge 2 commits intomainfrom
Conversation
…pipelines OSWorld and CL-bench both implement the same evolution pattern (propose skills after solve, curate per-topic, curate general patterns) but as standalone scripts without using the engine abstraction. This adds a shared ProposeCurateEngine that implements EvolutionEngine.step() with the common pipeline, and provides engine-based example scripts for both benchmarks. New files: - agent_evolve/algorithms/propose_curate/ — engine + prompts - examples/osworld_examples/evolve_osworld_engine.py - examples/cl_bench_examples/evolve_cl_bench_engine.py Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…al prompts - Rename prompts.py constants to DEFAULT_TOPIC_CURATOR_PROMPT/DEFAULT_GENERAL_CURATOR_PROMPT - Engine constructor now accepts topic_curator_prompt, general_curator_prompt, and format_failed_summary parameters for full customization - Rewrite evolve_osworld_engine.py with IDENTICAL original prompts and full helper functions (trajectory signals, compress, bot detection, eval text, build_propose_messages) to reproduce results exactly - Rewrite evolve_cl_bench_engine.py with IDENTICAL original prompts and full helper functions (rephrase feedback, in-context propose, parse proposal) - Both examples pass exact original curator prompts to engine constructor Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
ProposeCurateEngine(agent_evolve/algorithms/propose_curate/) — a sharedEvolutionEngineimplementation that encapsulates the propose+curate pipeline used by both OSWorld and CL-benchevolve_osworld_engine.py,evolve_cl_bench_engine.py) that use the new engine viaengine.step()agent_evolve/algorithms/__init__.pyalongside existing enginesThe engine handles:
feedback.raw["proposal"])Configurable via
skill_layoutparameter:"topic"(OSWorld),"context"(CL-bench), or"flat".Test plan
step()with empty observations returnsmutated=Falsefeedback.raw["proposal"]works🤖 Generated with Claude Code