Unverified50% confidenceBenchmarkExact time
TGIT's zero-shot transfer to real human instructions saw the success rate jump from 11.33% to 32.51%.
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter 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%.80% similarUnverifiedOn an unseen OpenFly test set, TGIT improved the success rate from 4.95% to 20.79%.70% similarUnverifiedIn GPT Researcher's experiments, the decision-based approach improved the proportion of effective context by approximately 27 percentage points compared to traditional embedding retrieval.63% 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%.61% similarUnverifiedMACTS-EM achieves a 22–27% improvement in zero-shot transfer capability.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/990451API
curl https://kongchang.com/api/v1/knowledge/claims/990451MCP
get_claim(id=990451)