Yitong Li

Lytttttt's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in xlangai/osworld_v2_tasks about 2 months ago

task_075: deliver assets via mirrored zip instead of sparse git clone (fix solution-leak hack)

#56 opened about 2 months ago by

Lytttttt

New activity in xlangai/osworld_v2_tasks 2 months ago

task_090: score only content that actually renders on the slide (fix off-canvas / pasted-image reward hack)

#54 opened 2 months ago by

Lytttttt

task_041: include BibTeX block in the instruction; drop un-instructed local-PDF check

#49 opened 2 months ago by

Lytttttt

task_008: make attachment vision descriptions discriminative between ticket receipts and boarding passes

#48 opened 2 months ago by

Lytttttt

task_089: score rendered layout via CDP instead of exact CSS literals; make requirement notes additive

#29 opened 2 months ago by

Lytttttt

task_090: fix nested-group shape flattening; align S14/S20 instruction with scoring

#28 opened 2 months ago by

Lytttttt

New activity in xlangai/osworld_v2_tasks 3 months ago

task_079: void VLM credit on slides faked with a full-slide picture background (anti-hacking)

#19 opened 3 months ago by

Lytttttt

task_056: serve degraded reference to agent + tighten MLT structure check (anti-hacking)

#18 opened 3 months ago by

Lytttttt

New activity in xlangai/osworld_v2_tasks 4 months ago

Fix evaluators: tasks 077 (L0 false positive) + 079/087/096 (L2 robustness)

#9 opened 4 months ago by

Lytttttt

New activity in xlangai/osworld2.0_human_crosscheck 5 months ago

Add task_004 eval.log (manual cross-check)

#1 opened 5 months ago by

Lytttttt

Add task_004 eval.log (manual cross-check)

#1 opened 5 months ago by

Lytttttt