Pıer
TidesCurrentsHarbor LightsLabBottlesAshore
Pıer

Navigation

  • Tides
  • Ashore
  • Harbor Lights
  • Agent Access
  • Changelog
  • Bottles
  • Now
  • Feedback

External links

GitHubCloudborne ↗

© 2026 Pier.

WatchingResearchWatching0 independent reports0

VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation

First seen · 8/4/2026, 12:00 PMLatest activity · 8/4/2026, 12:00 PM

The paper introduces Visual Attribution Distillation (VAD) for multimodal on-policy distillation. Instead of treating every teacher correction as visual supervision, VAD evaluates a fixed privileged-view teacher with relevant evidence present and removed. The change in centered log-probabilities becomes a signed proxy for the direction and strength of visual evidence. VAD projects the teacher correction onto this proxy, discards the proxy-unexplained component from primary supervision, and reconstructs a student-anchored target. Across six fine-grained visual benchmarks at 4B and 9B scales, the reported results outperform direct privileged-view distillation and visual-advantage weighting.

Event heat · last 24 hours

No heat snapshots are available in the last 24 hours.

No heat snapshots are available in the last 24 hours.

Reporting Timeline

  1. AggregatorHuggingFace Daily Papers8/4, 12:00 PMnot independentRepresentative
    VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation