Unverified50% confidenceFactExact time
After preference alignment on Qwen3.5-4B using length-normalized DPO plus an SFT retention term (the RPO objective), the performance change on the structured extraction task was -0.55 percentage points, passing the TOST equivalence test within the ±5 percentage point equivalence interval.
1
Sources
50%
Confidence
Long-term
Relevance
11/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn the Qwen3-VL-4B model, fine-tuning with procedurally generated data raises its accuracy on SpaceCast-Bench from 34.0% to 65.7%, and achieves a macro-average gain across six out-of-domain benchmarks.71% similarUnverifiedThe 4-bit quantization of Qwen3 27B shows robust performance, with the gap from FP16 baseline on mainstream benchmark tasks remaining within an acceptable range.71% similarUnverifiedIn evaluations across seven benchmarks, D3D achieves an average decoding speedup of 5.00x with Qwen3-8B as the target model.70% similarUnverifiedWhen run on Qwen3.6-35B-A3B with a 50% expert pruning ratio, SlimWise increases decoding throughput by up to 1.81x with minimal accuracy loss.70% similarUnverifiedIn the Qwen3-4B comparative experiment, updating the model directly with unfiltered skills without using SGUID filtering decreased performance by 0.3 percentage points on HMMT25.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1138283API
curl https://kongchang.com/api/v1/knowledge/claims/1138283MCP
get_claim(id=1138283)