Repository navigation

All pull requests

Labels (12)Milestones (0)New pull request

Open

(19)

Closed

(66)

Fix task annotation errors: typos, wrong reference answers, and incorrect eval values
#274zjordon opened Aug 27, 2026
fix(eval): check last line for LLM judge verdict in fuzzy_match and ua_match
#272AUTHENSOR opened Aug 5, 2026
docs: reference ClawBench live-web benchmark
#271reacher-z opened Aug 1, 2026
Support HuggingFace completion prompts for non-Llama models
#266mitre88 opened May 23, 2026
Replace deprecated judge model
#252allenz opened May 5, 2026
Add MiniMax as LLM provider for web agent evaluation
#250octo-patch opened Mar 24, 20263 of 4 tasks
correct README
#241Kotstantinovskiy opened Nov 28, 2025
add trajectory JSONL streaming for persistence/debug/replay
#240thanay-sisir opened Nov 27, 2025
add automatic 3-attempt retry for LLM parse failures
#239thanay-sisir opened Nov 27, 2025
add exact text match to TYPE action equivalence in is_equivalent
#238thanay-sisir opened Nov 26, 2025
Unify Consecutive Repeating Action Detection in early_stop
#237thanay-sisir opened Nov 26, 2025
prioritize recent obs content via suffix truncation
#236thanay-sisir opened Nov 26, 2025
Fix dropdown interaction bug with DOM manipulation
#229leejamesss opened Oct 15, 2025
2
Update test.raw.json
#176hage1005 opened Sep 17, 2024
Add missing __init__.py to llms/providers
#153zhudotexe opened Jun 28, 2024
Feature/fix pop menu
#112lwaekfjlk opened Mar 14, 2024Member
11
Wrong scroll action specification in the prompts vs. in the parsing function
#110michalspiegel opened Mar 13, 2024
2
Update evaluators.py
#77Miking98 opened Dec 20, 2023
Add RCI Agent
#38robert1003 opened Sep 12, 2023Collaborator

Navigation Menu

Repository navigation

All pull requests

Fix task annotation errors: typos, wrong reference answers, and incorrect eval values

fix(eval): check last line for LLM judge verdict in fuzzy_match and ua_match

docs: reference ClawBench live-web benchmark

Support HuggingFace completion prompts for non-Llama models

Replace deprecated judge model

Add MiniMax as LLM provider for web agent evaluation

correct README

add trajectory JSONL streaming for persistence/debug/replay

add automatic 3-attempt retry for LLM parse failures

add exact text match to TYPE action equivalence in is_equivalent

Unify Consecutive Repeating Action Detection in `early_stop`

prioritize recent obs content via suffix truncation

Fix dropdown interaction bug with DOM manipulation

Update test.raw.json

Add missing init.py to llms/providers

Feature/fix pop menu

Wrong scroll action specification in the prompts vs. in the parsing function

Update evaluators.py

Add RCI Agent

Pull requests