---
title: "Accuracy is not compatibility: two Evo 2 checkpoints hinge on one FP8 matrix multiplication"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2610.39415v1"
permalink: "https://feed.astralor.com/items/itm_86fb24acb05f4f97d49e7db7"
published: "2026-10-29T05:55:59.000Z"
score: 54
channel: "ai"
tags: ["Accuracy","Evo","FP8","DNA","Hopper FP8","Blackwell"]
---

# Accuracy is not compatibility: two Evo 2 checkpoints hinge on one FP8 matrix multiplication

## At a glance

FP8 formats specify how values are stored, not how products are summed; this gap can break checkpoints. Evo 2, a DNA language model, documents a Hopper FP8 requirement for its 1B and 40B checkpoints. On two Blackwell (s…

## Key points

- FP8 formats specify how values are stored, not how products are summed; this gap can break checkpoints.
- Evo 2, a DNA language model, documents a Hopper FP8 requirement for its 1B and 40B checkpoints.
- On two Blackwell (sm_120) GPUs both fail the developers' forward test (the 7B passes); on an RTX 5070 Ti the 40B's next-base accuracy is 25.7%, near guessing.

## Full text

FP8 formats specify how values are stored, not how products are summed; this gap can break checkpoints. Evo 2, a DNA language model, documents a Hopper FP8 requirement for its 1B and 40B checkpoints. On two Blackwell (sm_120) GPUs both fail the developers' forward test (the 7B passes); on an RTX 5070 Ti the 40B's next-base accuracy is 25.7%, near guessing. Greater accuracy breaks them, as BioNeMo saw for the 40B on Hopper: natively, 99.996% of one projection's outputs are correctly rounded (H100: about 30%), and fp32 accumulation of the same operands fails. Our H100 FP8 GEMM emulator, mmaemu (bit-exact on captured 1B projections), localises the dependence to the 1B's first projection (of 21), where exact sums fail but pass when shrunk by the H100's 1.4% (measured in-sample), not 2.1%. The 40B fails with only its first projection native. Emulating only that projection, weights unchanged, raises 40B BRCA1 variant-classification AUROC from 0.434 to 0.869 (H100: 0.868) on the RTX 5070 Ti, despite a narrow forward-test miss, and 0.481 to 0.867 on an RTX PRO 6000. We tested one model family but not training origin; releases should specify reference arithmetic.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_86fb24acb05f4f97d49e7db7