Unverified50% confidenceBenchmarkExact time
Fine-tuning a coding model on SWE-rebench data filtered with ExLabel, the model solved 66.2% of problems on SWE-bench Verified, 14.5% higher than the model trained on data filtered with the best baseline.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/30/2026
First Seen
Valid until: 11/13/2026
Sources
Related Entities
Related Claims
UnverifiedIn the Slop Code Bench benchmark, a certain frontier model scored only 14.8%.64% similarUnverifiedMeta's LLaMA paper disclosed that approximately 67% of its training data came from Common Crawl, of which only about 15% of the original data was retained after multiple rounds of filtering.64% similarUnverifiedAfter matched supervised training, model performance improved to 99.8%, while removing the relational links caused accuracy to plummet to 23.4%, even below the best shortcut's 27.0%.63% similarUnverifiedWhen sampling eight candidates without constraints from the model, 10.2% of inputs cannot produce any valid output under limited training data, and this proportion drops to 0.4% when using the full training data63% similarUnverifiedThe model is trained in two stages: first masked language model pre-training on about 20 million sentences from FineWeb-Edu, then correction fine-tuning on 12.02 million (noisy, clean) pairs, of which 30% are identity pairs.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1113920API
curl https://kongchang.com/api/v1/knowledge/claims/1113920MCP
get_claim(id=1113920)