REVIEW 22 cited by
ReasonIR: Training Retrievers for Reasoning Tasks
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
ReasonIR: Training Retrievers for Reasoning Tasks
read the original abstract
We present ReasonIR-8B, the first retriever specifically trained for general reasoning tasks. Existing retrievers have shown limited gains on reasoning tasks, in part because existing training datasets focus on short factual queries tied to documents that straightforwardly answer them. We develop a synthetic data generation pipeline that, for each document, our pipeline creates a challenging and relevant query, along with a plausibly related but ultimately unhelpful hard negative. By training on a mixture of our synthetic data and existing public data, ReasonIR-8B achieves a new state-of-the-art of 29.9 nDCG@10 without reranker and 36.9 nDCG@10 with reranker on BRIGHT, a widely-used reasoning-intensive information retrieval (IR) benchmark. When applied to RAG tasks, ReasonIR-8B improves MMLU and GPQA performance by 6.4% and 22.6% respectively, relative to the closed-book baseline, outperforming other retrievers and search engines. In addition, ReasonIR-8B uses test-time compute more effectively: on BRIGHT, its performance consistently increases with longer and more information-rich rewritten queries; it continues to outperform other retrievers when combined with an LLM reranker. Our training recipe is general and can be easily extended to future LLMs; to this end, we open-source our code, data, and model.
Forward citations
Cited by 22 Pith papers
-
FollowTable: A Benchmark for Instruction-Following Table Retrieval
FollowTable is the first large-scale benchmark for instruction-following table retrieval, paired with an Instruction Responsiveness Score, showing that existing models fail to adapt to fine-grained constraints beyond ...
-
LeanSearch v2: Global Premise Retrieval for Lean 4 Theorem Proving
LeanSearch v2 recovers 46.1% of ground-truth premise groups for research-level Lean 4 theorems within 10 candidates and raises fixed-loop proof success to 20%.
-
LeanSearch v2: Global Premise Retrieval for Lean 4 Theorem Proving
LeanSearch v2 recovers 46.1% of ground-truth premise groups on research-level Mathlib theorems and raises fixed-loop proof success from 4% to 20% via embedding-reranker plus iterative sketch-retrieve-reflect retrieval.
-
MARVEL: Multimodal Adaptive Reasoning-intensiVe Expand-rerank and retrievaL
MARVEL reaches 37.9 nDCG@10 on the MM-BRIGHT benchmark by combining LLM query expansion, a reasoning-enhanced dense retriever, and GPT-4o CoT reranking, beating prior multimodal encoders by 10.3 points.
-
Antares: Foundation Models for Agentic Vulnerability Localization
Antares-3B, a 3B model trained with SFT plus GRPO, matches GPT-5.5 on repository-scale vulnerability localization at roughly 1/100th the inference cost.
-
TabRank: Chain-of-Thought Distillation for Table Re-Rankers
Conditioning a table reranker on teacher chain-of-thought traces during training, without supervising generation of the traces, improves out-of-distribution table retrieval accuracy over base, naive SFT, and standard ...
-
A Survey of Reasoning-Intensive Retrieval: Progress and Challenges
A survey that categorizes RIR benchmarks by domain and modality, proposes a taxonomy for integrating reasoning into retrieval pipelines, and outlines key challenges.
-
An Agentic Approach to Metadata Reasoning
Metadata Reasoner uses agentic LLM reasoning on metadata to select sufficient and minimal data sources, achieving 83.16% F1 on KramaBench and 85.5% F1 on noisy synthetic benchmarks while avoiding low-quality tables 99...
-
HIVE: Query, Hypothesize, Verify An LLM Framework for Multimodal Reasoning-Intensive Retrieval
HIVE raises multimodal retrieval nDCG@10 to 41.7 on the MM-BRIGHT benchmark by inserting LLM-driven hypothesis generation and verification between retrieval passes, delivering +9.5 over the best text-only baseline and...
-
Are LLM-Based Retrievers Worth Their Cost? An Empirical Study of Efficiency, Robustness, and Reasoning Overhead
Empirical comparison across 14 retrievers on the BRIGHT benchmark shows reasoning-specialized models can match strong accuracy with competitive speed while many large LLM bi-encoders add latency for small gains and co...
-
Procedural Knowledge at Scale Improves Reasoning
Reasoning Memory decomposes reasoning trajectories into 32 million subquestion-subroutine pairs and retrieves them via in-thought prompts to improve language model performance on math, science, and coding benchmarks b...
-
Procedural Knowledge at Scale Improves Reasoning
Retrieving compact procedural hints from 32M subquestion–subroutine pairs improves reasoning-model accuracy on math, science, and coding benchmarks beyond compute-matched test-time scaling.
-
Revisiting Text Ranking in Deep Research
In deep-research agents, passage-level BM25 retrieval plus re-ranking beats larger neural rankers, and converting agent queries into natural questions (Q2Q) improves neural ranking.
-
With Argus Eyes: Assessing Retrieval Gaps via Uncertainty Scoring to Detect and Remedy Retrieval Blind Spots
An entity-level Retrieval Probability Score exposes systematic blind spots in dense retrievers, predicts them from embedding geometry, and targeted Wikipedia-context augmentation of flagged entities raises nDCG@5/10 o...
-
Benchmarking Information Retrieval Models on Complex Retrieval Tasks
CRUMB is a new benchmark for complex, multi-aspect retrieval tasks on which state-of-the-art retrieval models score poorly, and query rewriting does not rescue the best models.
-
ERank: Fusing Supervised Fine-Tuning and Reinforcement Learning for Effective and Efficient Text Reranking
ERank trains a pointwise reasoning-LLM reranker with fine-grained integer scoring and a listwise reinforcement-learning reward, reporting nDCG@10 of 40.2 on BRIGHT for its 32B variant with BM25 hybrid scoring.
-
SHIFT: Self-reconstruction Harnesses Implicit Fine-grained Thinking for Retrieval
SHIFT trains LLM retrievers to reason through latent continuous tokens and reconstruct explicit reasoning traces, improving reasoning-intensive retrieval.
-
Context Convergence Improves Answering Inferential Questions
Passages made from high-convergence sentences improve LLM performance on inferential questions compared to cosine similarity selection.
-
BRIDGE: Multimodal-to-Text Retrieval via Reinforcement-Learned Query Alignment
BRIDGE reaches 29.7 nDCG@10 on MM-BRIGHT by RL-aligning multimodal queries to text and using a reasoning retriever, beating multimodal encoders and, when combined with Nomic-Vision, exceeding the best text-only retrie...
-
Beyond Sequential Reranking: Reranker-Guided Search Improves Reasoning Intensive Retrieval
Reranker-Guided-Search, a greedy graph search steered by reranker scores, outperforms sequential top-k reranking under a fixed budget on three reasoning-intensive retrieval benchmarks.
-
Domain-Aware RAG: MoL-Enhanced RL for Efficient Training and Scalable Retrieval
A two-stage RAG training pipeline, MoL continual pre-training plus GRPO reinforcement learning with single-passage training and multi-passage inference, reports state-of-the-art retrieval recall on NFCORPUS and SCIFAC...
-
Granite Embedding Multilingual R2 Models
Granite Embedding Multilingual R2 releases 311M and 97M parameter bi-encoder models that achieve state-of-the-art retrieval performance on multilingual text, code, long-document, and reasoning datasets.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.