0.991 AUROC zero-shot on insertions/deletions

EVEE demonstrates strong generalization to indels without explicit training, indicating learned mechanistic principles.

Source paper

extracted_from

Explaining 4.2 million genetic variants with state-of-the-art, interpretable predictions

(2026) · Pearce, Michael · Dooms, Thomas · Yamamoto, Ryo · Meehl, Joshua +18

Neighborhood — ranked by edge-count

Papers (1)

paper

Explaining 4.2 million genetic variants with state-of-the-art, interpretable predictions
mentions

Claims (1)

claim

Evee outperforms existing methods for variant pathogenicity prediction
supports
Interpretive claim supported by the high AUROC findings.

Communities (2)

community

Mechanistic interpretability & model evaluation
members_of
Spans attention head decomposition, benchmark awareness, and genomic pathogenicity prediction via neural models.
Mechanistic explanation of genetic variant pathogenicity
members_of
Using genomic foundation model internals to generate disruption profiles that explain variant effects mechanistically, achieving 0.997 AUROC on ClinVar pathogenicity prediction.

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Reflection direction features achieve AUROC 0.772 vs. 0.736 for final layer baseline on deepseek-llama-8b on GSM8k correctness predictionfinding0.716
Supports claim that uncertainty is encoded in reflection direction
0.997 AUROC on pathogenicity prediction for 839k ClinVar variantsfinding0.716
EVEE achieves state-of-the-art performance on variant pathogenicity classification, outperforming existing methods.
System prompting and few-shot produce highly similar activations (cosine 0.83-0.92 at safety-critical layers), indicating a shared prompt-side mechanism distinct from activation steering.claim0.711
Evidence for two representational pathways based on cross-method activation divergence
P25 neutral baseline on DeepSeek-R1 shows 28.8% ASR under activation steering at alpha=4.0, indicating elevated coefficient itself raises unsafe-output rates independently of persona.finding0.708
Baseline AS vulnerability of DeepSeek-R1 at elevated coefficient
Predicting dataset correctness from change in steered cross-entropy loss yields AUPRC = 0.91 when averaging over 6000 promptsfinding0.708
Steered loss can identify whether a dataset is likely to lead to misalignment
F0-trained probes in layers 4-10 show inverted separation on F1 (AUROC ≈ 0), systematically misclassifying true statements as false.finding0.707
Demonstrates that early-layer probes capture sentence polarity rather than truth.
No-prompt probes show significant AUROC performance drop when evaluated on ask-correct activations, especially at layers where arithmetic truth directions emerge under no-prompt.finding0.704
Generalization evidence that truth probes are not invariant to model instructions.
F3-trained probes achieve AUROC ~0.6 on F4, showing generalization breakdown from counting over 2 to 5 cities.finding0.702
Demonstrates the sharp drop in factual truth generalization at the counting boundary.