Pıer
TidesCurrentsHarbor LightsLabBottlesAshore
Pıer

Navigation

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

External links

GitHubCloudborne ↗

© 2026 Pier.

WatchingResearchWatching0 independent reports0

Wnuan: Staged Post-Training for Question Answering over Proprietary Enterprise Knowledge

First seen · 8/3/2026, 04:00 AMLatest activity · 8/3/2026, 04:00 AM

Wnuan is a three-stage post-training pipeline for adapting models to proprietary enterprise knowledge: task-oriented supervision is generated from documents, supervised fine-tuning uses general-data replay, and reinforcement learning targets residual errors. According to the supplied abstract, the main 32B route improves acceptable-answer rate on the 707-question WnuanBench from 52.76% before adaptation to 80.06% after SFT and 91.51% after RL. The reported gain comes with a 5.17-point average decline on general benchmarks, concentrated in instruction following.

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 Papers8/3, 04:00 AMnot independentRepresentative
    Wnuan: Staged Post-Training for Question Answering over Proprietary Enterprise Knowledge