Datasets:
Use 70-task cohort across all results; select published GLM 200K results; separate errors and timeouts
a686cce verified Download data/missing-materials.csv from TIGER-Lab/WebsiteBenchArtifacts: direct link, hf CLI and curl.
- Browser
- Download file 515 Bytes
-
https://huggingface.co/datasets/TIGER-Lab/WebsiteBenchArtifacts/resolve/main/data/missing-materials.csv
- Command line
-
hf download hf://datasets/TIGER-Lab/WebsiteBenchArtifacts/data/missing-materials.csv
-
curl -L -o missing-materials.csv https://huggingface.co/datasets/TIGER-Lab/WebsiteBenchArtifacts/resolve/main/data/missing-materials.csv
515 Bytes
| task_id,task_name,model_id,model_name,trajectory_found,website_found,has_published_score,metadata_path | |
| etsy,Etsy,qwen3.8-max,Qwen 3.8 Max,True,False,True,benchmark/etsy/qwen3.8-max/index.json | |
| everyplate,EveryPlate,gpt-6.1-sol,GPT 6.1 Sol,True,False,True,benchmark/everyplate/gpt-6.1-sol/index.json | |
| simplify-jobs,Simplify Jobs,qwen3.8-max,Qwen 3.8 Max,True,False,True,benchmark/simplify-jobs/qwen3.8-max/index.json | |
| target,Target,claude-opus-5,Claude Opus 5,True,False,False,benchmark/target/claude-opus-5/index.json | |