← Back to all news
arXiv Machine Learning

Predictable Failure in Multi-Hop Retrieval: Score-Distributional Confidence Scoring and Abstention

The paper shows that multi‑hop retrieval failures cluster in predictable subpopulations and formalizes this with two theoretical results: (1) confident‑failure reduction is possible only when retrieval features carry mutual information about success, and (2) no single ANN score feature dominates across all failure regimes. Building on these insights, the authors introduce RegimeAbstain, which computes a Retrieval Confidence Score (RCS) from up to nine query‑ANN structural features and uses it to calibrate an abstention policy. Across three benchmarks and two retrieval architectures, RCS achieves the best or co‑best AUC‑AC and significantly reduces the Confident‑Wrong‑Answer Rate, demonstrating its effectiveness and domain‑agnostic applicability.

Related stories

arXiv Computation and Language
Aug 25

DynaKRAG: A Unified Framework for Learnable Evidence Control in Multi-Hop Retrieval-Augmented Generation

DynaKRAG is a unified framework that learns a state‑conditioned policy to control evidence acquisition in multi‑hop retrieval‑augmented generation. It uses a deterministic validity layer to build an action set, a learned continuation gate to decide between generating an answer or gathering more evidence, and an advantage scorer to rank evidence operations by predicted gain. Across HotpotQA, 2Wiki, and MuSiQue with various backbone models, DynaKRAG achieves top EM and F1 scores, improves token and retrieval efficiency, and enables terminal evidence compression that reduces context size while boosting answer quality.

By Chenyu Zhou, Yaqi Wu, Xiaolei Guo, Jiaqi Huang, Xianfa Zhang, Junxu Zhang, Zhuo Yu, Zhubo Shi, Jianghao Lin, Dongdong Ge
llmsrag
More like this →
arXiv AI
Aug 28

Evaluating Confidence-Gated Retrieval with Matched Trajectory Replay

The paper introduces matched trajectory replay, a protocol that fixes answer states, evidence points, budgets, and action costs to evaluate how confidence signals influence agent actions. Using this method, the authors compare raw verbalized confidence with post‑hoc isotonic calibration across six model‑dataset pairs, finding that calibration can significantly improve accuracy of committed answers but may reduce coverage and increase retrieval usage. The study concludes that calibration helps interpret commitment risk but does not predict the benefit of additional retrieval, indicating the need for separate value‑of‑information estimates.

By Prateek Chhikara
llmsagents
More like this →
arXiv AI
Aug 28

Why RAGs Hallucinate: Penalty-Aware Evaluation of Retrieval-Augmented Generation Systems with Knowledge-Gap Canaries

The paper introduces a penalty‑aware evaluation framework for Retrieval‑Augmented Generation (RAG) systems that uses asymmetric scoring, knowledge‑gap canaries, and a failure‑attribution pipeline. Applying this framework to three commercial RAG products and a baseline on SimpleQA‑Verified, the authors find that while overall accuracy is similar across systems, canary violation rates vary dramatically, showing that systems differ more in when they answer than in what they answer. The study demonstrates that penalty‑aware scoring can reorder system rankings and is robust across different penalty settings.

By Alden Do Rosario, Hussein Younes, Felipe Pires
rag
More like this →
arXiv AI
Aug 25

XTC: Head-Aware Sampling by Excluding Top Choices

XTC (Exclude Top Choices) is a lightweight, head‑aware decoding operator that improves diversity in autoregressive language models by removing overly probable tokens that dominate the next‑token distribution. It works by identifying tokens above a plausibility threshold, probabilistically excluding the dominant choices, and renormalizing the remaining distribution. Across 60 experiments on models such as Gemma 3 and DeepSeek R1, XTC boosts Distinct‑2 scores by 11–15 % and cuts repeat trigrams by 27–47 %, while a Mechanical Turk study shows a 62.3 % preference for XTC‑generated text without loss of fluency.

By Philipp Emanuel Weidmann, Allen Roush, Judah Goldfeder, Sanjay Basu, Ravid Shwartz-Ziv
llmsefficiency
More like this →
arXiv Computation and Language
Sep 23

ABAI at COLIEE 2026 Task 1: Multi-Stage Retrieval with GraphRAG-Enhanced Meta-Learning, and a Post-Hoc Study of the Cross-Validation-to-Test Gap

The paper reports the ABAI submission to COLIEE 2026 Task 1, a case law retrieval challenge that suppresses cited passages, and details a four‑stage retrieval pipeline: multi‑view BM25 with reciprocal rank fusion, neural reranking, graph‑based features via a graph attention network, and a LightGBM meta‑learner over 34 features. The best run achieved an F1 score of 0.177 on the official test set, compared to a cross‑validated 0.311, and the authors attribute the gap to a recall ceiling, temporal distribution shift, and threshold miscalibration. A controlled post‑hoc study examined the impact of threshold transfer, decision quality across time, and query similarity, and identified specific remedies—such as BM25 length‑normalisation tuning, event‑triple views, and dense fusion—that improved recall, while other interventions had no effect.

By Minhan Cho, Soyoung Park, Daejin Choi, Jinyoung Han
rag
More like this →