reranking

一篇实验性博客文章提出「Bag of Decisions」重排方法：让 LLM 为每个查询生成一组是/否判准问题，再用 Jev 计算各问题为真的概率并求和作为相关性分数，在 Wayfair WANDS 和 Amazon ESCI 数据集上均优于 BM25 和单一问题重排。

A benchmark comparing 18 RAG pipeline variants against a simple agent loop on Google's FRAMES multi-hop questions finds the best pipeline hits 78.9% while an agentic loop that can re-search reaches 92.7%, with surprising results showing reranking often hurts accuracy. The authors also note models sometimes fill gaps from memory despite instructions to stick to retrieved documents.

This paper introduces component-aware feedback for LLM-guided evolutionary search, logging which component edits changed which fitness metrics in an attribution memory, achieving baseline-quality performance in one-third of the search budget and 7.2% higher nDCG@10 on Bright reranking tasks.

介绍 Jev 如何作为 RAG 管道中的“裁判”，在检索与生成之间评估候选段落的相关性、可回答性和潜在提示注入，并附上基于 Comet Opik 的开源 Jev-as-a-Judge 项目。

RenderRank introduces a reranking method that uses compressed visual document representations to reduce token usage and improve efficiency while maintaining accuracy across multiple datasets.

The paper presents ZooWork-ShopRanker, an open-source family of e-commerce rerankers aligned to judge-labeled human preferences, achieving significant performance gains over baselines and releasing models and a new benchmark for research.

This paper presents a controlled study on learned context planning for long-context multiple-choice QA, demonstrating that it does not robustly outperform strong retrieval methods like BM25 and hybrid retrieval under various experimental settings.

This article tests the application of the Jev model in RAG retrieval, evaluating its effects on accelerating retrieval and reducing costs. Results show advantages in reranking and judging answerability.

The tweet recommends Jev for RAG, highlighting its superior semantic matching ability compared to dot product similarity, making it useful as a similarity metric and reranker.

A developer shares an experiment using Qwen 3.5 4B to mimic Jev's probability output by grabbing logit probabilities, with results and code available on GitHub and a demo website.

Alpine Agentic Search is an AI community meetup in Zurich on October 7, featuring technical talks on vector search, reranking, and agents for search workloads, with speakers from Nebius, NVIDIA, and Tavily.

OntologyAligner presents a three-stage framework combining ontology-aligned retrieval and LLM reranking for biomedical ontology normalization, achieving state-of-the-art performance on HPO tasks and introducing the PhenoNormBench benchmark.

This paper proposes Diverse Skill Routing (DSR), a diversity-aware reranking framework using Determinantal Point Processes to balance relevance and non-redundancy in skill selection for LLM agents, improving recall and coverage on benchmarks.

This paper introduces hLLM, a decoding strategy for generative reranking that uses the Hungarian algorithm to achieve single-pass decoding, resulting in a 64x speed-up while maintaining ranking quality.

This paper evaluates LLM rerankers in conversational recommendation systems, demonstrating that performance and stability are highly dependent on retrieval protocols, candidate pool configuration, and decoding settings.

The article introduces PACE, a training-free framework for optimizing RAG systems by reordering candidates based on evidence coverage and dynamically adjusting reranking budgets, improving evidence recall and reducing latency.

A discussion on the cost-effectiveness of embeddings versus LLMs, referencing a study called 'embedder's dilemma' that finds LLMs outperform embedding models at significantly higher cost.

The author shares techniques to prevent RAG-based agents from acting on bad retrieved context, including hybrid retrieval, reranking, corrective retrieval steps, and guardrails, with a build lab event on Aug 29 to demonstrate these methods.

This paper presents HNR-DAC, a two-stage framework for scientific claim verification over cited papers, combining hard-negative reranking and distribution-aligned classification. It achieves strong results on NLPCC 2026 Task 10 Track 2, ranking third on the leaderboard with the highest Macro-F1.

This paper introduces UniME-R1, an embedder-adviser framework for unified multimodal retrieval that generates Retrieval-Centric Chain-of-Thought (RC-CoT) conditioned on retrieval feedback, improving retrieval performance by learning from hard negatives.