WeFair: Fidelity-Aware Alignment from Human Feedback for Personalized Query Rewriting in Short-Video Search
READ THE ORIGINAL SOURCEOpen original sourcehuggingface.co · Opens in a new tabWeFair is a fidelity-aware alignment framework for personalized query rewriting in short-video search. It stages human feedback from demonstration (SFT) to preference (DPO, from subsequent consumption) to verification (GRPO, rewarding exact agreement with reformulation or preservation targets) to constrain intent drift. Offline, WeFair-2B outperforms strong rewriting baselines in overall accuracy and intent preservation. An online A/B test shows a 2.593% relative lift in effective video-view rate and a 2.350% relative reduction in query reformulation rate. The model is deployed in WeChat short-video search, which serves over 100 million queries per day.
One primary source; other links are context, not independent corroboration.
- Primary sourceHugging Face Daily Papers
- Discovery linkHugging Face
- Linked in sourceView arXiv page
- Linked in sourceView PDF
- Linked in sourcehttps://arxiv.org/abs/2610.16927v2
The paper page is linked. Peer review and findings have not been independently checked by AINewsIQ.
Open full article view →- Original paper+45
- Fetched source page+30
- No editor link review+0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
WeFair is a fidelity-aware alignment framework for personalized query rewriting in short-video search. It stages human feedback from demonstration (SFT) to preference (DPO, from subsequent consumption) to verification (GRPO, rewarding exact agreement with reformulation or preservation targets) to constrain intent drift. Offline, WeFair-2B outperforms strong rewriting baselines in overall accuracy and intent preservation. An online A/B test shows a 2.593% relative lift in effective video-view rate and a 2.350% relative reduction in query reformulation rate. The model is deployed in WeChat short-video search, which serves over 100 million queries per day.
A READER'S EVIDENCE CHECKHow to read this signalSix questions about claims, sources, risks, and open checks
This is a signal-level guide until claim-level evidence is available. “Not recorded” means the evidence has not been assessed here; it does not mean a result is false. How we check AI news →
- 01Who is making each claim?
The claimant and whether each claim was measured or asserted are not recorded in this public brief.
- 02What was actually tested?
Not recorded per claim. Check the source for sample size, tasks, model and benchmark versions, and reproducibility.
- 03Where are the sources?
Original link: huggingface.co. Brief basis: fetched source page. Primary page date and quote locations are not established in this panel.
- 04What could mislead?
Source claims not independently checked; the online gains are reported as relative percentages from the authors' own A/B test.
- 05What is at risk if wrong?
Not assessed per claim. Consider safety, privacy, cost, and adoption decisions before relying on the result.
- 06What would change the view?
No claim-specific check is recorded. An inspectable primary method or independent replication could change the assessment.
Key findings from the stored brief
- Strict, verifiable human feedback was found more effective than richer behavioral rewards for preventing intent drift.
- WeFair-2B outperforms strong rewriting baselines offline in overall accuracy and intent preservation.
- Online A/B testing showed a 2.593% relative lift in effective video-view rate and 2.350% relative reduction in reformulation rate.
Source claims not independently checked; the online gains are reported as relative percentages from the authors' own A/B test.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
The paper page is linked. Peer review and findings have not been independently checked by AINewsIQ.
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
A publication revision history is not stored for this signal. New engagement snapshots do not mean the source itself changed.
Why this may matter
Search systems that over-optimize engagement can drift from users' immediate intent. WeFair suggests grounding training in verifiable reformulations can keep compact rewriters faithful while still improving outcomes at large production scale.
Limits and open questions
Source claims not independently checked; the online gains are reported as relative percentages from the authors' own A/B test.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.