Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.17614 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912660864368640 |
|---|---|
| author | Singh, Praphul Barrett, Corey Srivasta, Sumana Bulu, Irfan Gadde, Sri Kenthapadi, Krishnaram |
| author_facet | Singh, Praphul Barrett, Corey Srivasta, Sumana Bulu, Irfan Gadde, Sri Kenthapadi, Krishnaram |
| contents | Clinicians need ranking systems that work in real time and still justify their choices. Motivated by the need for a low-latency, decoder-based reranker, we present OG-Rank, a single-decoder approach that pairs a pooled first-token scoring signal with an uncertainty-gated explanation step. The model scores all candidates in one pass and generates a brief, structured rationale only when the list is genuinely ambiguous, keeping latency predictable. Trained with a curriculum that concentrates effort on hard cases, OG-Rank delivers strong effectiveness on encounter-scoped order selection (fast path: Recall@1~0.45, nDCG@20~0.625) and improves further when the gate activates (Recall@1~0.56, nDCG@20~0.699 at a 45\% gate rate), while compact backbones show similar gains under the same policy. Encoder baselines trail in both effectiveness and flexibility. The result is a practical recipe: rank fast by default and explain when it helps, a pattern that applies broadly to decision tasks where selective generation buys accuracy at acceptable cost. The single-policy design simplifies deployment and budget planning, and the curriculum principle (spend more on the hard cases, less on the easy ones) readily transfers beyond clinical order selection. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_17614 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | OG-Rank: Learning to Rank Fast and Slow with Uncertainty and Reward-Trend Guided Adaptive Exploration Singh, Praphul Barrett, Corey Srivasta, Sumana Bulu, Irfan Gadde, Sri Kenthapadi, Krishnaram Artificial Intelligence Information Retrieval Clinicians need ranking systems that work in real time and still justify their choices. Motivated by the need for a low-latency, decoder-based reranker, we present OG-Rank, a single-decoder approach that pairs a pooled first-token scoring signal with an uncertainty-gated explanation step. The model scores all candidates in one pass and generates a brief, structured rationale only when the list is genuinely ambiguous, keeping latency predictable. Trained with a curriculum that concentrates effort on hard cases, OG-Rank delivers strong effectiveness on encounter-scoped order selection (fast path: Recall@1~0.45, nDCG@20~0.625) and improves further when the gate activates (Recall@1~0.56, nDCG@20~0.699 at a 45\% gate rate), while compact backbones show similar gains under the same policy. Encoder baselines trail in both effectiveness and flexibility. The result is a practical recipe: rank fast by default and explain when it helps, a pattern that applies broadly to decision tasks where selective generation buys accuracy at acceptable cost. The single-policy design simplifies deployment and budget planning, and the curriculum principle (spend more on the hard cases, less on the easy ones) readily transfers beyond clinical order selection. |
| title | OG-Rank: Learning to Rank Fast and Slow with Uncertainty and Reward-Trend Guided Adaptive Exploration |
| topic | Artificial Intelligence Information Retrieval |
| url | https://arxiv.org/abs/2510.17614 |