Unverified50% confidenceBenchmarkExact time
cAST's AST-based chunker achieves a 4.3 percentage point Recall@5 improvement on RepoEval retrieval and a 2.67 percentage point Pass@1 improvement on SWE-bench generation.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter re-evaluating with non-overlapping anchors, the net advantage instead grew from +4.1pp to +5.5pp, but significance dropped from 2.4 standard errors to 1.4 standard errors; the inflation factor is about 2.4x, exactly equal to sqrt(912/163).68% similarUnverifiedDeReAct improves the Pass@1 of Claude Sonnet 4.5 by 4.2 to 5.2 percentage points.66% similarUnverifiedAfter GRPO reinforcement learning, checkpoints using TailSFT achieved a final pass@1 improvement of up to 3.9 points.66% similarUnverifiedOn Terminal Bench, Fable 5.1's score is approximately 20 points higher than Grok 4.7's, representing close to a 50% improvement.64% similarVerifiedOn the DeepSuite benchmark, Grok 4.7 scored 71%, GPT 5.6 Sol scored 72.7%, Fable 5.1 scored 70%, and Astra Max topped the leaderboard at 74.1%.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1024140API
curl https://kongchang.com/api/v1/knowledge/claims/1024140MCP
get_claim(id=1024140)