Papers/2609.30275
🧪 Test?View on arXiv

Cosine Similarity Is Not Evidence: Measuring the Noise Floor of Interpretability Transfer Under Quantization

P. Varshney, A. Author, B. Author, C. Author, D. Author

quantizationinterpretabilityAI safetymodel evaluation
2609.30275
Builder Relevance
70%
1h ago

Abstract

The paper argues that reported statistics like cosine similarity are insufficient for interpreting the preservation of interpretability artifacts when transitioning from full-precision to quantized weights in AI models.

Reality Card

Core Claim

The study demonstrates that cosine similarity cannot be reliably interpreted without knowing the underlying class separation and sample size, leading to potential misinterpretations of model performance under quantization.

Method / Result

At INT4 quantization, the direction of the decision variable rotated significantly, exceeding the estimator's noise, while at INT8, no movement was detected.

Limitations

The main limitation is the lack of reported sample sizes in existing studies, which hinders the ability to assess the reliability of cosine similarity as a measure of preservation.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.
← Back to all papers