Unverified50% confidenceBenchmarkExact time
According to the author's report in the Reddit post, ORL increases the number of tokens per forward pass of QwenDLM-4B by up to 6.2 times, improves end-to-end throughput by about 6.9 times, and raises HumanEval pass@1 from 82.9% to 89.6%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/17/2026
First Seen
Valid until: 10/31/2026
Sources
Related Entities
Related Claims
UnverifiedAt a throughput of 259k, Qwen3.5 achieves approximately a 19× speed improvement, with hybrid attention contributing approximately 5×, MoE sparsification approximately 2×, and Multi-Token Prediction approximately 1.5×.71% similarUnverifiedThe three-level cache improved generation speed under Q4 weights from 1.83 tokens per second to 8.89, an improvement of nearly 5x.68% similarUnverifiedWhen the candidate token acceptance rate is high, speculative decoding can effectively output 2–4 tokens per forward pass, improving overall generation speed by 1.5x–3x.67% similarUnverifiedOn Qwen-Image-Edit, Rubric-CEPR improved ImgEdit from 4.36 to 4.60, a gain of +5.5%.66% similarUnverifiedRe-quantizing INT4 weights to NF4 subjects each weight to two rounds of rounding, and official measurements show this increases the perplexity of Qwen3-8B by 1.6%65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1049947API
curl https://kongchang.com/api/v1/knowledge/claims/1049947MCP
get_claim(id=1049947)