Unverified50% confidenceBenchmarkExact time
On the Qwen3-VL-8B model, the crossover of correct and incorrect trajectories yielded 272 correct descendant trajectories confirmed by the verifier out of 621 attempts, accounting for 43.8%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/4/2026
First Seen
Valid until: 2/2/2027
Sources
Related Entities
Related Claims
UnverifiedTesting with the Qwen model yielded consistent conclusions, with the token approach achieving 67.0% accuracy versus 64.8% for real data.71% similarUnverified3,189 successful rollouts were collected from Qwen3.8-27B for supervised fine-tuning (imitation learning), and performance stagnated at 47.5%, below the results of two rounds of RL training.69% similarUnverifiedIn the first round of tests (23 synthetic cases × 3 runs), qwen2.5:1.5b achieved only a 30% correct-judgment rate and qwen2.5:7b only 46%, both failing; qwen2.5:7b incorrectly judged "covered" in 14% of cases.69% similarUnverifiedIn the archived Qwen experiments, the study conducted a gradient writing test on Qwen3.8-Flash-Next-FP8, and the specified first answer token was ranked highest among 132 independently reset targets, of which 93 were newly promoted and 39 were already in first place.68% similarUnverifiedOn the Qwen-3.5-4B-Instruct model, the top two principal axes matched the style dimensions of human spontaneous requests with a precision of 72.8% and a macro-averaged recall of 43.6%.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133768API
curl https://kongchang.com/api/v1/knowledge/claims/1133768MCP
get_claim(id=1133768)