DecisionRank on-policy distillation beats generative teacher on retrieval tasks
Item-space OPD lets 1.7B DecisionRank surpass 27B LLM teacher on TREC DL while running 24.5x faster.
Stories per day · 30 days
2 RSS
Relatedretrieval3closed-set1deberta1decision-models1embeddings1encoders1generator1ms-marco1nemotron1nvidia1on-policy-distillation1rag1tool-selection1
13 stories, newest first · one card per story
Stories tagged #reranking
Glance reranks passages from the generator's shared prefill
A resident 9B backbone adds about 0.1 GiB per skill and beats bge-reranker-v2-m3 on SciFact and NFCorpus after 39 minutes of fine-tuning.
Per-option encoders beat joint scoring on closed-set decisions
At 32 million parameters, per-option scoring gains 5.8 skill points but needs 5 to 13 times more passes.
NVIDIA announces Nemotron 3.5 Embed, Rerank, and Parse models
The post body was unavailable, so sizes, benchmarks, licenses, and release status are not stated.
Evaluating Biomedical Reranking for LLM-Based Question Answering over Longitudinal Clinical Notes
Biomedical reranking raised clinical-note Hit@10 from 46.6% to 60.6%, while Qwen3-8B correctness reached only 48.6%.
Decision-Oriented Recommendation Reranking: An Empirical Study of Jev
TypeSafe AI's Jev holds recommendation quality while latency grows more slowly than pointwise Qwen rerankers.
Decision-Oriented Recommendation Reranking: An Empirical Study of Jev
Jev matches strong recommendation rerankers with slower latency growth than pointwise Qwen models.
RenderRank: Learning to Rerank Text with Compressed Visual Tokens
RenderRank reranks documents from compressed visual tokens, using 16.5-35.5% fewer tokens and beating sub-4B text baselines.
When Does Selection Replace Extraction? A Pre-Registered Test of Agent Memory with a Typed Decision Model
Pre-registered tests show raw turns selected by Jev match extracted agent memory at far lower cost.
Equal Ranking Quality, Different Decisions: Measuring and Reducing Order Dependence in LLM Scorers
Order-consistency fine-tuning reduces LLM scorer decision changes from candidate reordering without hurting ranking quality.
Think Thrice Before Reranking: Multi-perspective Evidence and Reasoning Integration for Text Reranking
MERIT-Rank integrates multiple LLM reasoning trajectories for text reranking; its 4B model outperforms most 7B and 32B rerankers on BRIGHT.
Generative Late-Interaction Embeddings For Visual Document Retrieval
GLIE compresses visual document retrieval embeddings to four vectors per page while retaining nearly 80% of uncompressed nDCG@5 accuracy.
Beyond Top-k Skill Retrieval: Diversity-Aware Skill Routing for LLM Agents
DSR reranks LLM agent skills with Determinantal Point Processes to favor complementary, non-redundant sets, improving multi-skill query coverage.