SLIDE Distills Beam Search into List-Wise Targets for Single-Rollout Generative Retrieval
_reachsumit · x · 2026-10-28
Generative retrieval typically relies on item-level beam search, leaving cross-item dependencies unmodeled and list-level objectives unoptimized; directly trained list-wise models underperform beam search under greedy decoding due to beam search's broader search space and positional bias in scarce positive signals.
SLIDE decouples offline search from online list generation: a teacher retrieves high-quality candidates via offline beam search, and their Semantic IDs are concatenated in beam-score order as list-wise distillation targets. The student then generates the full Top-K list in a single greedy pass, decoupling teacher scaling.
More from Research
- TTME beats test-time training on 5 math tasks while cutting FLOPs by 82.8-84.6% — UCSantaBarbara · 2026-10-28
- EnvLunch turns executable environments into free verifiable agent training data — KingsCollegeLondon · 2026-10-28
- Anthropic reveals it is developing its own experimental drugs using Claude — Polymarket · 2026-10-28
- Stanford launches TaskNet: 132 worker interviews build a large-scale occupational task inventory — soumitrashukla9 · 2026-10-28
- KAIST's Principia uses LLMs to construct and refine scientific principles at test time — coallaoh · 2026-10-28
- The AlphaFold Data Problem: Quantum Chemistry Models as Training Data for Larger Scales — TimothyDuignan · 2026-10-28