arXiv AI
Jul 23
Information Discernment in Large Language Models
arXiv:2607. 19355v1 Announce Type: new Abstract: LLMs are increasingly used with external knowledge sources like the internet.
Models in the Same Family are NOT Trust-Equivalent
Related stories
arXiv AI
Aug 11
Embedding Trust: Semantic Isotropy Predicts Nonfactuality in Long-Form Text Generation
arXiv:2510. 21891v2 Announce Type: replace-cross Abstract: To deploy large language models (LLMs) in high-stakes application domains that require substantively accurate responses to open-ended prompts, we need reliable, computationally inexpensive methods that assess the trustworthiness of long-form responses generated by LLMs.
arXiv AI
4d ago
The Unequal Influence of Bad Advice: Using Training Data Attribution to Modulate Emergent Misalignment
arXiv:2609.37914v1 Announce Type: cross Abstract: Fine-tuning large language models on narrow, misaligned tasks can undo their post-training alignment and induce novel misaligned behaviors -- a pheno...
arXiv Machine Learning
Jun 26
The Geometry of Updates: Fisher Alignment at Vocabulary Scale
arXiv:2606. 27242v1 Announce Type: new Abstract: Training-free source selection for LLM families with shared vocabularies arises in scientific string domains such as SMILES, protein, and genomic sequences, where candidate corpora share a tokenizer but differ in prediction targets.
arXiv Computation and Language
Sep 1
When Calibration Rankings Reverse: Accuracy-Controlled Evaluation for Fair Comparison of LLMs
The paper argues that traditional global calibration metrics, such as Expected Calibration Error and Brier Score, are confounded by differences in model accuracy when comparing large language models. It introduces ACE, an accuracy‑controlled evaluation framework that offers Instance‑Aligned, Distribution‑Aligned, and Candidate‑Aligned views to provide fairer cross‑model comparisons. Experiments across various benchmarks reveal that many reported calibration advantages disappear after accuracy control and that model rankings often reverse, indicating that raw global metrics are unreliable for cross‑model calibration assessment.
arXiv AI
Sep 15
Ventriloquist LLMs: Linear Alignment of Late-Stage Representations
arXiv:2603.18908v5 Announce Type: replace Abstract: Independently trained language models often learn compatible late-stage representations, despite differences in training objectives, architectures,...
arXiv AI
4d ago
PADM\'E: Preference Alignment Data Synthesis for Meta-Evaluation of LM Agent Evaluators
PADM'E is a method for synthesizing preference‑aligned data to meta‑evaluate language‑model (LM) evaluators of agentic behaviors. It reframes meta‑evaluation as a preference judgment problem, generating criterion‑based data with small LMs and no human involvement. In a prototype, PADM'E produced 1,000 samples across four domains and three criteria, and human validation showed agreement with human judgment rising from 73% to 85% compared to a naive baseline.
arXiv AI
Sep 25
Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models
The study investigates whether large language models (LLMs) can identify code they have generated, potentially leading to self‑favoring or collusive behavior. Experiments across 15 model‑benchmark pairs show that models can attribute authorship with balanced accuracy between 49% and 58%, but this ability largely stems from superficial cues such as solution length. Removing surface features like docstrings, comments, and type hints reduces attribution accuracy to chance, indicating that surface cues drive the effect.
arXiv Machine Learning
Aug 10
Latent Fact-Checking: Detecting Misinformation through Activation Engineering
arXiv:2608. 06417v1 Announce Type: new Abstract: The proliferation of misinformation online has driven demand for scalable detection systems.
arXiv AI
Sep 15
How Semantically Stable Are LLM Refusals? Measuring Confusion in Local Safety Boundaries
The paper introduces Semantic Confusion to assess how consistently large language models refuse similar prompts. It presents ParaGuard, a 10k‑prompt corpus of controlled paraphrase clusters, and proposes three token‑level metrics—Confusion Index, Confusion Rate, and Confusion Depth—to measure contradictory refusal decisions across meaning‑preserving paraphrases. Experiments show that global false rejection rates can mask local inconsistencies, revealing that refusal evaluation must consider both frequency and consistency across nearby paraphrases.
arXiv Machine Learning
Aug 20
Stability-Aware Feature Design for Robust Watermark Detection in Machine-Generated Text
The paper introduces Pattern Stability Score (PSS), a watermark detection framework that uses local statistical features and stability dynamics across paraphrased variants to identify machine-generated text. PSS combines global and local z‑score features with higher‑order run‑length statistics, autocorrelation signals, and stability scores over paraphrase depth. Experiments on PG‑19, CNN/DailyMail, and WikiText with Llama‑3‑8B, Qwen2‑7B, and multiple paraphrasers show that PSS improves detection AUC by 10‑15 percentage points and a single universal classifier achieves over 87.8% AUC across diverse LLMs, paraphrasers, and domains without retraining.
arXiv AI
Jul 28
Beyond Shapley: An Influence-Based Data Auditing Pipeline for LLM Alignment and Evaluation
arXiv:2607. 22766v1 Announce Type: cross Abstract: The alignment of Large Language Models (LLMs) is increasingly bottlenecked by data quality.