Unverified50% confidenceBenchmarkExact time
After TGIT training, the success rate for weak instruction input increased from 15.27% to 37.93%, surpassing the original exhaustive command rate of 31.03%.
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTGIT's zero-shot transfer to real human instructions saw the success rate jump from 11.33% to 32.51%.80% similarUnverifiedUsing only synthetic weak instructions fed directly into the navigator, the success rate was 15.27%, falling between the exhaustive command rate of 31.03% and the real user instruction rate of 11.33%.75% similarUnverifiedAfter matched supervised training, model performance improved to 99.8%, while removing the relational links caused accuracy to plummet to 23.4%, even below the best shortcut's 27.0%.68% similarUnverifiedOn an unseen OpenFly test set, TGIT improved the success rate from 4.95% to 20.79%.67% similarUnverifiedIn certain settings, students with stronger prior ability advance at a rate up to 3.26 times higher than students with weaker prior ability.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/990450API
curl https://kongchang.com/api/v1/knowledge/claims/990450MCP
get_claim(id=990450)