REVIEW 20 cited by
A Stitch in Time Saves Nine: Detecting and Mitigating Hallucinations of LLMs by Validating Low-Confidence Generation
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
read the original abstract
Recently developed large language models have achieved remarkable success in generating fluent and coherent text. However, these models often tend to 'hallucinate' which critically hampers their reliability. In this work, we address this crucial problem and propose an approach that actively detects and mitigates hallucinations during the generation process. Specifically, we first identify the candidates of potential hallucination leveraging the model's logit output values, check their correctness through a validation procedure, mitigate the detected hallucinations, and then continue with the generation process. Through extensive experiments with GPT-3.5 (text-davinci-003) on the 'article generation task', we first demonstrate the individual efficacy of our detection and mitigation techniques. Specifically, the detection technique achieves a recall of ~88% and the mitigation technique successfully mitigates 57.6% of the correctly detected hallucinations. Importantly, our mitigation technique does not introduce new hallucinations even in the case of incorrectly detected hallucinations, i.e., false positives. Then, we show that the proposed active detection and mitigation approach successfully reduces the hallucinations of the GPT-3.5 model from 47.5% to 14.5% on average. We further demonstrate the effectiveness and wide applicability of our approach through additional studies including performance on different types of questions (multi-hop and false premise questions) and with another LLM from a different model family (Vicuna). In summary, our work contributes to improving the reliability and trustworthiness of large language models, a crucial step en route to enabling their widespread adoption in real-world applications.
Forward citations
Cited by 20 Pith papers
-
Before Agents Speak: Pre-hoc Failure Risk Inference in Multi-Agent Systems
HalluProp infers per-agent and system-level hallucination risk in multi-agent LLMs before interaction via role–query misalignment, topology-aware propagation, and differentiable Noisy-OR aggregation.
-
DICA: Dual-Indicator Guided Contrastive Alignment in Multimodal Large Language Models
Monitoring attention entropy and image-output correlation during decoding, then applying targeted contrastive corrections, reduces hallucination in multimodal LLMs without retraining.
-
Reasoning Denoiser: Denoising Reasoning Traces for Hallucination Detection in Large Reasoning Models
REDE removes irrelevant and repetitive reasoning steps using final-answer attention as supervision, improving hallucination detection AUROC by up to ~19 points over unfiltered traces.
-
When Should LLMs Be Less Specific? Selective Abstraction for Reliable Long-Form Text Generation
Atom-wise selective abstraction—replacing low-confidence factual claims with higher-confidence, less specific versions—improves the risk-coverage trade-off in long-form generation by up to 27.73% AURC over claim removal.
-
Geometry-Aware Hallucination Detection in Large Language Models
A manifold-based prototype sampling method for choosing in-context examples improves hallucination-detection accuracy over several ICL baselines in the majority of tested settings.
-
Neural Message-Passing on Attention Graphs for Hallucination Detection
CHARM trains graph neural networks on token-attention graphs built from LLM computational traces and outperforms prior hallucination detectors on five benchmarks at token and response level.
-
MIRAGE-Bench: LLM Agent is Hallucinating and Where to Find Them
A new benchmark with a three-way hallucination taxonomy, snapshot-based test cases, and an LLM judge shows LLM agents hallucinate at over 30% of risky decision points, with open and closed models closer than expected.
-
Hallucination Detection and Mitigation with Diffusion in Multi-Variate Time-Series Foundation Models
Pre-trained multivariate time-series imputation models frequently return values that violate known relations between variables, and a diffusion-based score can detect and filter these errors.
-
The Curious Case of Factuality Finetuning: Models' Internal Beliefs Can Improve Factuality
Fine-tuning on model-generated text filtered by internal probes improves factual accuracy more than fine-tuning on gold documents across three long-form generation domains.
-
BugRepro: Enhancing Android Bug Reproduction with Domain-Specific Knowledge Integration
BugRepro combines RAG with UI transition graphs to improve LLM-based Android bug reproduction, reporting 96/151 success versus 55/151 for AdbGPT, though the evaluation may be contaminated by data overlap.
-
Pierce the Mists, Greet the Sky: Decipher Knowledge Overshadowing via Knowledge Circuit Analysis
PhantomCircuit traces knowledge overshadowing to attention circuits during training and prunes circuit edges to recover the overshadowed answer.
-
Not All Needles Are Found: How Fact Distribution and Don't Make It Up Prompts Shape Retrieval, Reasoning, and Hallucination in Long-Context LLMs
On a new extended needle-in-a-haystack benchmark, explicit anti-hallucination prompts and dispersed fact placement cause some long-context LLMs to over-refuse or collapse in accuracy, while others remain robust.
-
MALM: A Multi-Information Adapter for Large Language Models to Mitigate Hallucination
A graph-based adapter that connects input, context, and knowledge tokens reduces hallucination in LLM question answering and improves RAG performance across several benchmarks.
-
Maximizing Confidence Alone Improves Reasoning
Reinforcement learning with a reward equal to the negative entropy of the model's own token distribution improves benchmark accuracy without any labeled answers.
-
Paying Alignment Tax with Contrastive Learning
A contrastive learning framework with positive and negative example pairs improves faithfulness and slightly reduces toxicity on Reddit TL;DR summarization, but the central claim of avoiding the alignment tax is not e...
-
Beyond ROUGE: N-Gram Subspace Features for LLM Hallucination Detection
Singular values of label-grouped n-gram frequency tensors are used as MLP features for hallucination detection, with reported gains on HaluEval that rely on label-aware grouping.
-
FActBench: A Benchmark for Fine-grained Automatic Evaluation of LLM-Generated Text in the Medical Domain
Unanimous voting between NLI and chain-of-thought fact-checking yields scores closest to medical expert judgments on three of four tasks in the new FActBench benchmark.
-
Counterfactual Probing for Hallucination Detection and Mitigation in Large Language Models
Counterfactual Probing detects LLM hallucinations by measuring how much a model's confidence changes when a claim is altered to a plausible but incorrect variant, then hedges flagged statements with template-based mit...
-
The impact of fine tuning in LLaMA on hallucinations for named entity extraction in legal documentation
A vector-search segmenter plus fine-tuned LLaMA models outperforms regex for legal NER, with LoRA fine-tuning sharply reducing hallucinations.
-
Building Task Bots with Self-learning for Enhanced Adaptability, Extensibility, and Factuality
A thesis that combines self-learning from dialog logs, schema-guided prompting, and self-aligned factuality to build task bots with minimal human intervention.
Discussion (0). Continue with ORCID to comment.