Unverified50% confidenceFactExact time
In the experiment, Astra modified the weights of the small Qwen model, raising the MATH500 score from 7.9% to 48.1%, while the share of answers hitting the trial-run limit fell from 93% to 33%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/4/2026
First Seen
Valid until: 2/2/2027
Sources
Related Entities
Related Claims
UnverifiedIn a math repair experiment, a small open-source model scored only 7.9% on the Math 500 test; Astra raised the score to 48.1% by locating and subtracting a direction within two layers (about 130,000 weights, an edit ratio of 0.016%), and its instruction-following ability also improved roughly twofold.79% similarUnverifiedWithout any training, Astra modified only 0.016% of the weights of a small model, raising its score on the Math 500 benchmark from 7.9% to 48.1%.78% similarUnverifiedThe "Astra fixes Qwen's math ability" experiment used Python and an evaluation loop to find a steering vector that changed only a tiny fraction of the weights, raising the math benchmark score from single-digit percentage points to nearly half.75% similarUnverifiedAt the 1B scale, AsymLoop's average zero-shot accuracy on seven common-sense reasoning benchmarks improves from 49.58% to 50.85% (compared with Full Loop).74% similarUnverifiedAfter 50 practice runs, Qwen's capped answers dropped from 93% to 33%, the math score rose from 7.9 to 48.1, elementary word problems rose from 14 to 67, and the instruction-following test rose from 21 to 62; when given only 10 or 25 practice runs, the loop barely faded and the math score remained low.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1135436API
curl https://kongchang.com/api/v1/knowledge/claims/1135436MCP
get_claim(id=1135436)