YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Rank-5 SVD Reconstruction โ€” Qwen2.5-Math-1.5B (RELEX rebuttal artifact)

Oracle rank-5 reconstruction: theta_recon = theta_0 + V_5 V_5^T (theta_step - theta_0), V_5 = top-5 right singular vectors of the full 500-step observed delta trajectory (--mode reconstruct --rank 5). Base = Qwen/Qwen2.5-Math-1.5B. Subfolders step100/ .. step500/.

Eval config

TianHongZXY/MATH test, greedy temp=0.0, top_p=0.95, top_k=-1, num_generation=1, max_tokens=4096 (5000 problems). Accuracy = mean of the boolean label field.

Reference (same config)

RLVR real: 67.96/70.14/70.78/71.38/71.54 Rank-1 recon: 67.20/70.62/70.68/71.54/71.06 (relex-rlvr/Rank1-Reconstruction-Qwen2.5-Math-1.5B) Expected rank-5 ~= rank-1 -> components 2-5 add ~nothing with oracle coefficients.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support