Unverified50% confidenceBenchmarkExact time
训练完成后，Needle在30个测试样本上取得了96.7%的F1调用得分
1
Sources
50%
Confidence
Long-term
Relevance
7/6/2026
First Seen
Sources
Needle: Fine-Tuning a 26M Ultralight Tool-Calling Model Locally — Full Walkthrough
bilibili薛定猫AI7/3/2026
Related Claims
UnverifiedIn guided mode, the average end-to-end accuracy increased from 28.6% to 33.7%, and the ambiguity detection F1 score increased from 45.3% to 64.9%70% similarUnverifiedQwen 3.6 Max in neutral mode has an ambiguity detection F1 of only 16% and asks an average of only 0.07 questions per task, but once it does ask for clarification, the accuracy reaches 94.7%66% similarUnverifiedIn mixed-precision training, each FP16 parameter occupies 2 bytes of GPU memory, and each FP32 parameter occupies 4 bytes of GPU memory.60% similarUnverifiedBullet 在 SWE-bench Verified 基准测试中取得 95.8% 的成绩，位列前三，平均每个任务耗时 119 秒57% similarUnverifiedAccording to the MedViT paper, the model achieved a Macro F1 score of 0.7791 on ChestX-ray1455% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/112997API
curl https://kongchang.com/api/v1/knowledge/claims/112997MCP
get_claim(id=112997)