OpenEnv hackathon expert-in-the-loop track w/ AWM#428

Closed

sfc-gh-mhidayetoglu wants to merge 18 commits into

huggingface:mainfrom

sfc-gh-mhidayetoglu:add-agent-world-model

Commits on Mar 7, 2026

Commits on Mar 8, 2026

Add expert-advised agent runner and bug fixes for AWM environment
sfc-gh-mhidayetoglu
committed
Merge pull request #1 from sfc-gh-mhidayetoglu/expert-enhancement
sfc-gh-mhidayetoglu
authored
Add EXPERT_ENHANCEMENT.md with results, architecture, and learnings
sfc-gh-mhidayetoglu
committed
Add dynamic expert-in-the-loop runner for AWM environment
sfc-gh-kganesan
committed
Merge pull request #2 from sfc-gh-mhidayetoglu/dynamic-expert-in-the-loop
sfc-gh-kganesan
authored
Remove static expert and simplify to dynamic expert only
sfc-gh-kganesan
committed
Remove stale static expert references from dynamic expert docstring
sfc-gh-kganesan
committed
Merge pull request #3 from sfc-gh-mhidayetoglu/dynamic-expert-in-the-loop
sfc-gh-kganesan
authored
Add GRPO training results, reward graphs, and research directions to EXPERT_ENHANCEMENT.md
sfc-gh-kganesan
committed
Update EXPERT_ENHANCEMENT.md with actual adaptive training results
sfc-gh-kganesan
committed
Regenerate all charts to reflect adaptive 6E/2S strategy
sfc-gh-kganesan
committed
Update results with Try-First v2 strategy: +0.589 reward, 32% completion rate
sfc-gh-kganesan
committed
Update all README sections with 29-step results: +0.824 reward, 44.5% completion, all 3 phase transitions
sfc-gh-kganesan
committed
Fix architecture diagram and workflow to reflect Try-First strategy
sfc-gh-kganesan
committed
Restructure README as presentation with environment/task details and 4 new diagrams
sfc-gh-kganesan
committed
Add minimal GRPO training script and Colab notebook for hackathon
sfc-gh-kganesan
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

OpenEnv hackathon expert-in-the-loop track w/ AWM#428

OpenEnv hackathon expert-in-the-loop track w/ AWM#428
sfc-gh-mhidayetoglu wants to merge 18 commits into
huggingface:mainfrom
sfc-gh-mhidayetoglu:add-agent-world-model

Commits on Mar 7, 2026

Add Agent World Model environment (AWM-1K)

Replace debug print() with logger.info()

Commits on Mar 8, 2026

Add expert-advised agent runner and bug fixes for AWM environment

Merge pull request #1 from sfc-gh-mhidayetoglu/expert-enhancement

Add EXPERT_ENHANCEMENT.md with results, architecture, and learnings

Add dynamic expert-in-the-loop runner for AWM environment

Merge pull request #2 from sfc-gh-mhidayetoglu/dynamic-expert-in-the-loop

Remove static expert and simplify to dynamic expert only

Remove stale static expert references from dynamic expert docstring

Merge pull request #3 from sfc-gh-mhidayetoglu/dynamic-expert-in-the-loop

Add GRPO training results, reward graphs, and research directions to EXPERT_ENHANCEMENT.md

Update EXPERT_ENHANCEMENT.md with actual adaptive training results

Regenerate all charts to reflect adaptive 6E/2S strategy

Update results with Try-First v2 strategy: +0.589 reward, 32% completion rate

Update all README sections with 29-step results: +0.824 reward, 44.5% completion, all 3 phase transitions

Fix architecture diagram and workflow to reflect Try-First strategy

Restructure README as presentation with environment/task details and 4 new diagrams

Add minimal GRPO training script and Colab notebook for hackathon