Dataset Viewer
Auto-converted to Parquet Duplicate
type
stringclasses
2 values
harness
stringclasses
1 value
id
stringclasses
1 value
name
stringclasses
1 value
message
dict
session
trackio
agent_run
Agent repro run 2026-08-03 00:17 UTC
null
message
null
null
null
{ "role": "system", "content": "Event", "timestamp": 1785716260532 }
message
null
null
null
{ "role": "assistant", "content": "[Claim 0 fix — Motive computes motion-specific influence via detection pipeline]\n# Claim 0 fix — Motive computes motion-specific influence via detection pipeline\n\n## Claim (unredacted from verdicts.json)\nMotive computes motion-specific influence by detecting motion, forming\nm...
message
null
null
null
{ "role": "assistant", "content": "[Claim 2 fix — 74.1% preference win rate for Motive-selected fine-tuning]\n# Claim 2 fix — 74.1% preference win rate for Motive-selected fine-tuning\n\n## Claim (unredacted from verdicts.json)\nHuman pairwise evaluation reports a 74.1% preference win rate for Motive-selected\nfine...
message
null
null
null
{ "role": "assistant", "content": "[Claim 3 fix — Frame-length normalization removes attribution bias]\n# Claim 3 fix — Frame-length normalization removes attribution bias\n\n## Claim (unredacted from verdicts.json)\nFrame-length normalization prevents attribution rankings from being biased\ntoward longer clips and...
message
null
null
null
{ "role": "assistant", "content": "[Claim 4 fix — Influence scores are not merely high-motion clips]\n# Claim 4 fix — Influence scores are not merely high-motion clips\n\n## Claim (unredacted from verdicts.json)\nMotive's influence scores are not merely selecting high-motion clips;\ninfluential clips are those pred...
message
null
null
null
{ "role": "system", "content": "Event", "timestamp": 1785716260532 }

Agent traces

Agent sessions published from a Trackio Logbook.

Downloads last month
27