Unverified50% confidenceBenchmarkExact time
On the Qwen3.5-2B model, GAD-RL achieved a Micro Recall of 59.92% on CHAOS-Bench, surpassing GRPO and fixed-weight GRPO+OPD by 8.45 and 4.43 percentage points respectively
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/1/2026
First Seen
Valid until: 12/30/2026
Sources
Related Entities
Related Claims
UnverifiedGAD-RL achieved an Overall total score of 91.18 on OmniDocBench v1.666% similarUnverifiedOllama分发的DeepSeek R1各版本均经过GGUF格式量化处理（通常为Q4_K_M或Q8_0精度），将每个参数从16位浮点数压缩至4-8位，显存占用减少50%-75%，推理质量损失通常在5%以内65% similarUnverifiedThe UP host stated that the Qwen3 coding model scored approximately 61.7 on SWE-Bench Pro, while the top-tier closed-source model used as a reference scored 53.4 (this data contains inconsistencies and is for reference only).64% similarUnverifiedVia --quantization quark_mxfp4, SGLang dequantizes NVFP4 weights and requantizes them to MXFP4 at load time, achieving GSM8K accuracy recovery of 97.5%–100.2% on models such as MiniMax-M2.7 and GLM-5.1.63% similarVerifiedQwen3.8-Flash-Next achieved scores of 58.7 and 62.5 on the SWE-bench 1.1 and SWE-bench Pro evaluations, respectively.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/963914API
curl https://kongchang.com/api/v1/knowledge/claims/963914MCP
get_claim(id=963914)