Unverified50% confidenceBenchmarkExact time
In the experiment with the smallest Qwen 3.5 version (about 800 million weights), Astra modified 64 columns of numbers in each of layers 15 and 19 (about 130,000 weights), raising the standard math test score from under 8% to 48% and elementary word problems from 14% to 67%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/4/2026
First Seen
Valid until: 2/2/2027
Sources
Related Entities
Related Claims
UnverifiedBy editing an extremely small portion of the weights, Astra raised QUEN 3.5 0.8b's solve rate on math problems from roughly 1 correct out of every 13 problems to close to half.74% similarUnverifiedExperience internalization fine-tuning improved Qwen3-8B-27B's one-shot score from 37.07 to 48.22, while reducing generated tokens from 6.41 million to 570,000 (an 11-fold reduction).68% similarUnverifiedQwen3 4B-instruct scored 93% on math tasks under the normal-style prompt, dropping sharply to 43% when switched to the florid style.67% similarUnverifiedOnly 8.5% of Gemma 3 270M's answers hit the trial-run limit, whereas 93% of the small Qwen model's answers hit the limit.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1135408API
curl https://kongchang.com/api/v1/knowledge/claims/1135408MCP
get_claim(id=1135408)