Unverified50% confidenceBenchmarkExact time
On self-generated trajectories, JET reduces the number of tool calls by 11% to 33% while largely maintaining task scores.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/18/2026
First Seen
Valid until: 2/16/2027
Sources
Related Entities
Related Claims
UnverifiedIn comparison with SFT, JET achieves 18% to 28% fewer tool calls than SFT on expert trajectories while retaining most of SFT's score improvement.67% similarUnverifiedExperimental results show that RAFT reduces training input by 35.23% on average, and by up to 97.11%.67% similarUnverifiedBy providing a small reward for completing tasks in fewer steps, the number of tool calls was reduced by 31%.65% similarUnverifiedPilotech 2.0's official testing shows that its mechanisms cut overall task cost by 83%.63% similarUnverifiedRelative to brute-force search, the fine-tuned generator required 22 times fewer candidate evaluations in the generation step, while the off-the-shelf generator required 600 times fewer.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1182986API
curl https://kongchang.com/api/v1/knowledge/claims/1182986MCP
get_claim(id=1182986)