Oriel — AI & technology worth noticing, every day.

PathoJev: Efficient Pathology Decisions from Shared Visual Context

Paper · https://arxiv.org/abs/2610.40752

arXiv:2610.40752v1 Announce Type: new Abstract: Pathology workflows require multiple visual judgments about the same slide, from identifying tumor type to assessing grade and invasion. Generating and parsing a separate textual response for each structured question adds latency between visual evidence and usable decisions. Building on the System One paradigm, we introduce PathoJev, a pathology model that directly predicts answer probabilities for whole-slide and region-level assessment. PathoJev encodes each slide once and evaluates independent question suffixes in batches over the shared visual context. Reusing the hybrid backbone's recurrent states further limits the computation that grows with slide-context length for each additional question. Across three whole-slide benchmarks, the 9B and 27B models achieve mean accuracies of 77.2% and 78.1%, exceeding SlideChat by 5.4 and 6.3 percentage points. Both scales also maintain competitive accuracy on five ROI benchmarks. On WSI-Bench, the 9B model runs 9.1x faster than SlideChat. Shared context accelerates eight-question workloads by 6.9x and 5.9x over separate prefix encoding, rising to 11.2x and 8.9x for 64 questions. These results provide a reusable probabilistic interface for combining multiple visual judgments in whole-slide and local tissue assessment. Code: https://github.com/ChyaZhang/PathoJev.

First seen by Oriel on .

AI summary

PathoJev is a pathology model that reuses shared slide visual context to predict multiple structured judgments efficiently, improving accuracy and substantially reducing inference time.

  • PathoJev encodes each whole slide once, then batches independent question suffixes over the shared visual context to produce answer probabilities.
  • Reusing the hybrid backbone’s recurrent states reduces computation that would otherwise grow with slide-context length for every additional question.
  • Across three whole-slide benchmarks, the 9B and 27B models reached 77.2% and 78.1% mean accuracy, surpassing SlideChat by 5.4 and 6.3 percentage points.
  • On WSI-Bench, the 9B model was 9.1x faster than SlideChat, while shared context accelerated 64-question workloads by 11.2x and 8.9x.
  • Both model sizes retained competitive accuracy across five region-of-interest benchmarks, supporting whole-slide and local tissue assessment.

#pathology #medical imaging #vision-language models #inference optimization

Generated by gpt-5.6-sol. Check the original for details.

Aliases

SourceExternal IDURL
arXiv AI Papersoai:arXiv.org:2610.40752v1https://arxiv.org/abs/2610.40752 Read discussion