cs.LGJan 23, 2026

What Does a Cosine of 0.8 Mean? Monotone Calibration of Embedding Similarity

Authors: Nicolas Tacheny

Organizations: University of Mons (Mons, Belgium)

Abstract

Text embedding models turn sentences into vectors, and the cosine between two vectors is the usual measure of their similarity. Search, semantic caching and deduplication often accept a pair when this cosine reaches a fixed threshold. When people judge one pair of sentences more similar than another, the cosine usually agrees, but its value means different things for different models, because each model packs its scores into its own narrow range. The same threshold can therefore keep most pairs with one model and few with another. We propose to calibrate the score instead of changing the model. An increasing function fitted on sentence pairs rated by people translates the cosine of each model into the human rating scale and never reverses the order of two pairs. We fit three such functions for six models, ranging from a simple S-shaped curve to a free step function. On new pairs, calibrated scores agree with human ratings and the models agree with one another. One threshold selects pairs of similar quality whatever the model. The scale obtained belongs to the human ratings used to fit it, so its values shift on two datasets rated differently while the models stay comparable.

Figures & tables

Explore similar work

Jun 28, 2026

Anisotropy Decides Cosine vs. Rank Metrics for Text Embeddings

The standard way to compare two text embeddings is cosine similarity. Scattered studies report that a different metric does better, but never pin down the geometric condition that decides when, or why. We settle both with a comprehensive empirical study: nineteen parameter-free similarity metrics on nineteen encoders, from compact sentence transformers up to seven-billion-parameter large language models, across seven datasets. The answer is geometric. When an encoder spreads its variance evenly across directions, cosine is the best parameter-free choice and no other metric helps by a usable margin. When the variance concentrates into a few dominant directions, a property known as anisotropy, rank-based and L1-type metrics beat cosine by a clear margin. The absolute gain is modest, but because cosine starts low on these encoders it is a sizable relative improvement, around twenty percent on average and largest where cosine is weakest. What decides this is the geometry of the embedding space, not how the model was trained: where the two disagree, the metric follows the geometry. One number, the fraction of variance held by the single most dominant dimension, predicts how much the alternatives help across all nineteen encoders, with a rank correlation of 0.86 and a linear correlation of 0.95. To test this as the cause rather than a correlate, we project out the dominant directions: cosine recovers and the advantage of the other metrics nearly vanishes, but only on the encoders that were anisotropic to begin with. The effect is directional, not magnitude based, since it survives normalizing every vector to unit length. Among parameter-free metrics, then, cosine is the right tool wherever an encoder is well spread, which includes the fine-tuned embedders commonly deployed for retrieval, and we give a one-number diagnostic for when it is not.
Mar 31, 2026

What Cosine Similarity of Label Representations Can and Cannot Tell us

Cosine similarity is often used to measure the similarity of vector representations of neural network models. However, the cosine similarity of representations is not guaranteed to tell us anything about model probabilities. In this paper we show that for a softmax classifier, be it an image classifier or an autoregressive language model, the cosine similarity between label representations (called unembeddings in the paper) does not give any information on the probabilities assigned by the model. Specifically, we prove that given two unembeddings, it is possible to create another model which assigns the same probabilities for all inputs, but where the cosine similarity between the representations is now either 1 or -1. We also show that for a sigmoid classifier (where each input can be assigned multiple labels), all pairwise cosine similarities between the unembeddings define the set of possible label combinations. However, for softmax classifiers (where each input is assigned a ranking of the labels from most to least likely), we need all pairwise cosine similarities between all differences of unembeddings to know which rankings the model can predict. We conclude that it is misleading to interpret the cosine similarity between unembeddings without reference to the classifier that produced them.
Aug 6, 2026

Mapping Similarity Spaces across Embedding Models with Synthetic Query Probing

Retrieval-Augmented Generation systems rely on similarity scores to retrieve relevant content, yet scores are not directly comparable across embedding models due to differing geometric properties, complicating model migration and limiting threshold reuse. We study how similarity scores can be related by learning mappings between score distributions rather than embeddings. We introduce Synthetic Query Probing, generating queries from documents to create controlled query-chunk pairs, enabling large-scale, reference-free analysis of cross-model similarity behavior. We evaluate the approach on multiple embedding configurations and learn score conversion functions using linear, isotonic, and quantile mappings. Experiments on SciFact and a proprietary corpus show that while models largely agree on rankings, their absolute scores exhibit systematic distortions. Learned mappings partially align these spaces and improve threshold portability, with isotonic regression performing best. Our results highlight the need for cross-model calibration and position Synthetic Query Probing as a scalable framework for analyzing embedding comparability.