Image de couverture de livre

Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Hanna, Michael, Pezzelle, Sandro, Belinkov, Yonatan
Format:	Preprint
Publié:	2024
Sujets:	Machine Learning Computation and Language I.2.7
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Are formal and functional linguistic mechanisms dissociated in language models?
par: Hanna, Michael, et autres
Publié: (2025)

Position-aware Automatic Circuit Discovery
par: Haklay, Tal, et autres
Publié: (2025)

CRISP: Persistent Concept Unlearning via Sparse Autoencoders
par: Ashuach, Tomer, et autres
Publié: (2025)

Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMs
par: Nikankin, Yaniv, et autres
Publié: (2025)

REVS: Unlearning Sensitive Information in Language Models via Rank Editing in the Vocabulary Space
par: Ashuach, Tomer, et autres
Publié: (2024)

Large Language Models Generate Harmful Content Using a Distinct, Unified Mechanism
par: Orgad, Hadas, et autres
Publié: (2026)

AtManRL: Towards Faithful Reasoning via Differentiable Attention Saliency
par: Höth, Max Henning, et autres
Publié: (2026)

Do Models Know Why They Changed Their Mind? Interpretability and Faithfulness of Chain-of-Thought Under Knowledge Conflict
par: Venkata, Pruthvinath Jeripity
Publié: (2026)

Distinguishing Ignorance from Error in LLM Hallucinations
par: Simhi, Adi, et autres
Publié: (2024)

Constructing Benchmarks and Interventions for Combating Hallucinations in LLMs
par: Simhi, Adi, et autres
Publié: (2024)

Aligning Large Language Models for Faithful Integrity Against Opposing Argument
par: Zhao, Yong, et autres
Publié: (2025)

A Dataset for Metaphor Detection in Early Medieval Hebrew Poetry
par: Toker, Michael, et autres
Publié: (2024)

On Measuring Faithfulness or Self-consistency of Natural Language Explanations
par: Parcalabescu, Letitia, et autres
Publié: (2023)

Trust Me, I'm Wrong: LLMs Hallucinate with Certainty Despite Knowing the Answer
par: Simhi, Adi, et autres
Publié: (2025)

Where Should LoRA Go? Component-Type Placement in Hybrid Language Models
par: Borobia, Hector, et autres
Publié: (2026)

Masked by Consensus: Disentangling Privileged Knowledge in LLM Correctness
par: Ashuach, Tomer, et autres
Publié: (2026)

Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics
par: Nikankin, Yaniv, et autres
Publié: (2024)

ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs
par: Simhi, Adi, et autres
Publié: (2025)

Scaling Laws for Forgetting When Fine-Tuning Large Language Models
par: Kalajdzievski, Damjan
Publié: (2024)

Reasoning Models Know What's Important, and Encode It in Their Activations
par: Nikankin, Yaniv, et autres
Publié: (2026)

When Models Can't Follow: Testing Instruction Adherence Across 256 LLMs
par: Young, Richard J., et autres
Publié: (2025)

Evaluating the Efficacy of Hybrid Deep Learning Models in Distinguishing AI-Generated Text
par: Oketunji, Abiodun Finbarrs
Publié: (2023)

Pre-trained Models Perform the Best When Token Distributions Follow Zipf's Law
par: He, Yanjin, et autres
Publié: (2025)

Correctness is not Faithfulness in RAG Attributions
par: Wallat, Jonas, et autres
Publié: (2024)

Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework
par: Gu, Hao, et autres
Publié: (2025)

Verbosity Tradeoffs and the Impact of Scale on the Faithfulness of LLM Self-Explanations
par: Siegel, Noah Y., et autres
Publié: (2025)

Self-Pruned Key-Value Attention: Learning When to Write by Predicting Future Utility
par: Szilvasy, Gergely, et autres
Publié: (2026)

ReFACT: Updating Text-to-Image Models by Editing the Text Encoder
par: Arad, Dana, et autres
Publié: (2023)

Engineering A Large Language Model From Scratch
par: Oketunji, Abiodun Finbarrs
Publié: (2024)

Large Language Model (LLM) Bias Index -- LLMBI
par: Oketunji, Abiodun Finbarrs, et autres
Publié: (2023)

A Closer Look at Bias and Chain-of-Thought Faithfulness of Large (Vision) Language Models
par: Balasubramanian, Sriram, et autres
Publié: (2025)

Old Habits Die Hard: How Conversational History Geometrically Traps LLMs
par: Simhi, Adi, et autres
Publié: (2026)

HACK: Hallucinations Along Certainty and Knowledge Axes
par: Simhi, Adi, et autres
Publié: (2025)

Language Models Are Implicitly Continuous
par: Marro, Samuele, et autres
Publié: (2025)

Bayesian Attention Mechanism: A Probabilistic Framework for Positional Encoding and Context Length Extrapolation
par: Bianchessi, Arthur S., et autres
Publié: (2025)

Ensemble Language Models for Multilingual Sentiment Analysis
par: Hasan, Md Arid
Publié: (2024)

On the Effect of (Near) Duplicate Subwords in Language Modelling
par: Schäfer, Anton, et autres
Publié: (2024)

RBCorr: Response Bias Correction in Language Models
par: Bhatt, Om, et autres
Publié: (2026)

Compact Example-Based Explanations for Language Models
par: Schoenegger, Loris, et autres
Publié: (2026)

The Open Source Advantage in Large Language Models (LLMs)
par: Manchanda, Jiya, et autres
Publié: (2024)