Unverified50% confidenceBenchmarkExact time
Under matched batch-size-1 decoding latency, LRCC simultaneously improves both perplexity and downstream accuracy on Llama-3.2-1B.
1
Sources
50%
Confidence
Long-term
Relevance
10/8/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在27B混合架构、4B稠密模型、36B MoE三类模型上以8k上下文测试，Prefill阶段性能无变化，Decode阶段打补丁后快约1.4%62% similarUnverifiedAccording to the SWE-Bench Verified benchmark, Qwen 2.5 Coder 32B achieves a coding ability score of 61.759% similarUnverifiedThe H3 model is trained with resolutions processed as multiples of 32; keeping Multiple at 32 during inference avoids artifacts or performance degradation caused by size mismatches.58% similarUnverified千问3.8 27B的代码方向进步明显大于文本创作，剧本拆解能力相比3.6有提升但幅度有限58% similarUnverifiedThe evaluator found that GLM 5.2 remains strong in pure coding, but HY3 outperforms it in tasks involving aesthetics, physics feel, and simulation texture.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/986706API
curl https://kongchang.com/api/v1/knowledge/claims/986706MCP
get_claim(id=986706)