Tagged articles
hallucination detection
13 articles · Page 1 of 1
This comprehensive guide details a production-ready evaluation framework for AI Agents covering five essential test categories — drift, hallucination, privilege escalation, latency, and cost — with deterministic assertions, LLM-as-judge integration, CI/CD pipelines, and regression loops to prevent silent failures.
AI agent evaluationCI/CD Integrationcost monitoring
0 likes · 101 min read

This survey introduces the UIPCA diagnostic framework to dissect reasoning hallucinations across five dimensions — unsupported premises, invalid inferences, dependency reuse, visible consistency, and action-policy failures — and maps detection signals to concrete interventions while exposing gaps in causal evidence for reasoning–reliability trade-offs.
Causal InferenceUIPCA frameworkevidence-based AI
0 likes · 39 min read

HKUST researchers introduce LexAgentHallu, a hierarchical benchmark that evaluates hallucinations in legal AI agents by tracing entire reasoning trajectories, revealing that even top-performing systems exhibit hallucinations in 89% of execution traces and 68% of correct answers contain flawed reasoning.
Agent EvaluationHKUSTLegal AI
0 likes · 14 min read

This paper introduces DynHD, a hallucination detection method for diffusion large language models that analyzes denoising dynamics across token and temporal dimensions, achieving 84% AUROC on LLaDA-8B and Dream-7B without repeated sampling.
Denoising DynamicsDreamDynHD
0 likes · 8 min read


AntTechSep 15, 2026 · Artificial Intelligence
SingProbe introduces endogenous runtime guardrails that integrate risk detection directly into LLM inference by reusing hidden states, achieving comparable safety and hallucination detection to external guardrails while adding less than 0.5% decode overhead across 29 open-source models.
LLM guardrailsSGLangSingProbe
0 likes · 25 min read

Contra Labs' Design Crit dataset reveals that while AI can generate images, current AI judges barely outperform random guessing in assessing design quality, but a small fine‑tuned model can close nearly half the gap to human agreement by learning from expert‑rated criteria.
AI design evaluationDesign Crit datasetGenerative AI
0 likes · 16 min read

The article surveys the 2026 open‑source ecosystem for AI testing, detailing programmable runtimes, AI‑specific quality dimensions, testing‑as‑code practices, observability integration, real‑world case studies, and remaining challenges such as multimodal support and long‑context stability.
AI TestingDevOpsLLM
0 likes · 8 min read

The article examines four frequent mistakes when testing large language models—misusing functional coverage, conflating hallucination detection with fact‑checking, ignoring multi‑turn interaction decay, and relying on traditional performance metrics—while offering concrete verification methods, tools, and real‑world results to improve AI quality assurance.
AI quality assuranceLLM testingcognitive SLA
0 likes · 8 min read

This article explains why simple prompt‑based citation is insufficient for Retrieval‑Augmented Generation, introduces a sentence‑level attribution pipeline, combines semantic similarity with NLI verification, and presents practical hallucination detection and structured JSON output to ensure answer reliability.
LLM reliabilityNLIPrompt engineering
0 likes · 10 min read

QuCo‑RAG introduces a dynamic retrieval‑augmented generation framework that quantifies uncertainty using pre‑training corpus statistics, replacing unreliable model confidence with objective frequency and co‑occurrence evidence, achieving millisecond‑level hallucination detection, superior multi‑hop QA performance, and cross‑model transferability across various LLMs.
Dynamic RetrievalLLMRetrieval-Augmented Generation
0 likes · 9 min read

Apple’s new paper presents RL4HS, a reinforcement‑learning framework that uses span‑level rewards and class‑aware policy optimization to detect hallucinated text spans in large language models, outperforming GPT‑5 and other baselines and offering more precise, auditable error identification.
RL4HSReinforcement Learninghallucination detection
0 likes · 9 min read

This article explains the nature, types, and causes of hallucinations in large language models used for NL2SQL, reviews both unsupervised and supervised detection methods, and introduces an efficient token‑confidence based Active Sampling Detection (ASD) approach with practical deployment examples and future research directions.
AI safetyASDLLM
0 likes · 19 min read

The authors present a semantic‑graph‑enhanced uncertainty modeling framework that captures token, sentence, and paragraph dependencies, propagates uncertainty through entity relations and contradiction probabilities, and achieves roughly a 20 % gain in paragraph‑level hallucination detection on WikiBio and NoteSum compared with existing uncertainty‑based baselines.
Sentence-level ModelingToken-level ModelingUncertainty Modeling
0 likes · 13 min read
