# Evo2

Evo2 turns harness edits and their execution outcomes into training signal for the model that writes them, lifting SWE-bench Verified resolution from the strongest baseline's 41.0% to 48.3% and Polyglot pass@1 from 62.6% to 82.1%.

Evo2 turns harness edits and their execution outcomes into training signal for the model that writes them, lifting SWE-bench Verified resolution from the strongest baseline's 41.0% to 48.3% and Polyglot pass@1 from 62.6% to 82.1%.

At branching points in unitig graphs, de novo assembly previously had to choose a continuation from topology, paired-end evidence, or coverage heuristics, while the sequence itself was rarely scored directly. RASTRO scores each candidate extension by bits-per-base surprisal, either from a sample-specific n-gram model or from the pretrained DNA language model Evo2, and checks the choice against boundary k-mers counted in the raw reads. On HIV-1 and SARS-CoV-2 unitig graphs, the sample-adaptive RASTRO-S improved agreement with read-supported choices as context length grew from n=5 to n=10, above random selection and a longest-extension greedy baseline; the abstract and the results table disagree on the HIV-1 value at n=10.

At branching points in unitig graphs, de novo assembly previously had to choose a continuation from topology, paired-end evidence, or coverage heuristics, while the sequence itself was rarely scored directly. RASTRO scores each candidate extension by bits-per-base surprisal, either from a sample-specific n-gram model or from the pretrained DNA language model Evo2, and checks the choice against boundary k-mers counted in the raw reads. On HIV-1 and SARS-CoV-2 unitig graphs, the sample-adaptive RASTRO-S improved agreement with read-supported choices as context length grew from n=5 to n=10, above random selection and a longest-extension greedy baseline; the abstract and the results table disagree on the HIV-1 value at n=10.

EvSpark is a speculative decoding system for hybrid convolutional-recurrent-attention DNA foundation models such as Evo2: it verifies draft blocks in parallel and restores all three classes of inference state by selecting retained intermediate states without replay, reaching 2.96x on 43 real-sequence prompts and 3.27x including five synthetic controls on an Evo2 7B 48-prompt benchmark with three training seeds, retaining 1.84x-2.43x at 262k context, and yielding 2.18x-2.46x on real sequences and 2.51x-2.78x on the full suite for 20B and 40B targets.

EvSpark is a speculative decoding system for hybrid convolutional-recurrent-attention DNA foundation models such as Evo2: it verifies draft blocks in parallel and restores all three classes of inference state by selecting retained intermediate states without replay, reaching 2.96x on 43 real-sequence prompts and 3.27x including five synthetic controls on an Evo2 7B 48-prompt benchmark with three training seeds, retaining 1.84x-2.43x at 262k context, and yielding 2.18x-2.46x on real sequences and 2.51x-2.78x on the full suite for 20B and 40B targets.