Unverified50% confidenceBenchmarkExact time
After using TailSFT, the pass@16 metric on coding tasks during the SFT phase showed an absolute improvement of up to 16.8 points.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/7/2026
First Seen
Valid until: 12/6/2026
Sources
Related Entities
Related Claims
UnverifiedAfter GRPO reinforcement learning, checkpoints using TailSFT achieved a final pass@1 improvement of up to 3.9 points.71% similarUnverifiedDGM automatically more than doubled baseline performance on the SWE-bench software engineering benchmark, achieving an absolute improvement of 30 percentage points63% similarUnverifiedCompared to baseline methods, Dude achieves improvements of up to 22.8% in recall and precision.59% similarUnverifiedOn BU Bench V1, the 4B-parameter SFT model surpassed the 45% score reported by DeepSeek V4 Pro and Kimi K2.6 under the open-source Browser Use framework.57% similarUnverifiedNVFP4量化方案相比BF16基线实现了1.5倍的推理加速，同时在top-1准确率上保留了92%到97%的水平54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/873216API
curl https://kongchang.com/api/v1/knowledge/claims/873216MCP
get_claim(id=873216)