Unverified50% confidenceBenchmarkExact time
Compared with the approach of "fine-tuning every sub-step", AdaRSI achieved a slightly higher success rate while consuming only 28% of the GPU time and 22% of the self-collected demonstration data.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
Unverifiedgpt-oss-120b spent about 25 times the cost per correct answer on the distractor counting task, while its accuracy was 7.5 percentage points lower (not significant).73% similarUnverifiedThe statistical scope of the 21% efficiency improvement figure is not clearly specified and may be a composite score of code output volume, feature delivery speed, or engineers' subjective assessments.68% 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.67% similarUnverifiedPrompt优先方案（用一条超长指令串起数据分析流程）实测成功率不到40%67% similarUnverifiedBy selecting the appropriate model for different tasks, the solution rate improved by more than 23 percentage points, and the per-task cost was reduced to less than one-third of the original.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1032862API
curl https://kongchang.com/api/v1/knowledge/claims/1032862MCP
get_claim(id=1032862)