Search papers, labs, and topics across Lattice.
This paper critiques the reliance on cosine similarity in biomedical language models for causal inference, revealing that models like BioBERT and PubMedBERT inaccurately associate unrelated cross-domain pairs with high similarity scores. The authors propose a two-step method to enhance causal discovery by employing a contrastive pass and a hard negative mining approach, significantly improving the model's ability to differentiate between related and unrelated events. Key results show that their method increases correlation scores and discrimination metrics while achieving substantial gains in processing speed on optimized hardware.
False proximity in biomedical embeddings can mislead causal reasoning, but a new method boosts discrimination accuracy by over 2.30x while slashing query latency by 133x.
Ask a pretrained biomedical language model whether"cortisol 28 ug/dL"and"stock-market volatility"are related, and it returns a cosine similarity of 0.83 on a scale where 1.0 means identical. The two share no mechanism. This is not a corner case: every off-the-shelf biomedical encoder we tested (BioBERT, PubMedBERT, BioM-ELECTRA) scores unrelated cross-domain pairs between 0.76 and 0.92 when the answer should be near zero. Accuracy on cross-domain discrimination is 0%. Retrieval systems survive this, because a language model downstream filters the noise. A Large Behavioural Model (LBM), a foundation model whose subject is a person rather than a sentence, does not: it reasons over a graph of a user's life and treats embedding proximity as evidence that two events are causally linked. False proximity writes a false causal edge, and everything downstream inherits the error. Here, embedding geometry is not a tuning knob; it is correctness. We report the fix. A contrastive pass over 72,034 pairs raises PubMedBERT BIOSSES correlation from 0.633 to 0.828 and within-vs-across-domain separation from 1.05x to 1.63x. A second pass, BODHI, mines hard negatives from edges absent in a biomedical knowledge graph and lifts separation to 2.30x and the discrimination gap to +0.392, at a 4.5% BIOSSES cost. On an Intel Xeon 6737P with AMX, OpenVINO cuts single-query latency from 1367 ms to 10 ms (133x) and reaches 555 sentences/sec. One finding contradicts standard advice: FP16 beats INT8 on this silicon at every serving batch size, and we explain why. The same model on a no-AMX Ice Lake instance runs 13-27x slower. We release the benchmark suite, training corpora, the BODHI generator, and the OpenVINO scripts.
