Dissecting the Ullman Variations with a SCALPEL: Why do LLMs fail at Trivial Alterations to the False Belief Task?

2024-06-20 · Zhiqiang Pi, Annapurna Vadaparty, Benjamin K. Bergen, Cameron R. Jones

Recent empirical results have sparked a debate about whether or not Large Language Models (LLMs) are capable of Theory of Mind (ToM). While some have found LLMs to be successful on ToM evaluations such as the False Belief task (Kosinski, 2023), others have argued that LLMs solve these tasks by exploiting spurious correlations -- not representing beliefs -- since they fail on trivial alterations to these tasks (Ullman, 2023). In this paper, we introduce SCALPEL: a technique to generate targeted modifications for False Belief tasks to test different specific hypotheses about why LLMs fail. We find that modifications which make explicit common inferences -- such as that looking at a transparent object implies recognizing its contents -- preserve LLMs' performance. This suggests that LLMs' failures on modified ToM tasks could result from a lack of more general commonsense reasoning, rather than a failure to represent mental states. We argue that SCALPEL could be helpful for explaining LLM successes and failures in other cases.

📄 PDF Abstract BibTeX arXiv:2406.14737

Code (0)

등록된 구현이 없습니다.

PrivacyScalpel: Enhancing LLM Privacy via Interpretable Feature Intervention with Sparse Autoencoders

2025-03-14 · Ahmed Frikha, Muhammad Reza Ar Razi, Krishna Kanth Nakka, Ricardo Mendes 외

Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language processing but also pose significant privacy risks by memorizing and leaking Personally Identifiable Information (PII). Existing …

Memorization Privacy Preserving

Scalpel: Fine-Grained Alignment of Attention Activation Manifolds via Mixture Gaussian Bridges to Mitigate Multimodal Hallucination

2026-02-10 · Ziqiang Shi, Rujie Liu, Shanshan Yu, Satoshi Munakata 외 arxiv

Rapid progress in large vision-language models (LVLMs) has achieved unprecedented performance in vision-language tasks. However, due to the strong prior of large language models (LLMs) and misaligned attention across mod…

SCALPEL: Selective Capability Ablation via Low-rank Parameter Editing for Large Language Model Interpretability Analysis

2026-01-12 · Zihao Fu, Xufeng Duan, Zhenguang G. Cai arxiv

Large language models excel across diverse domains, yet their deployment in healthcare, legal systems, and autonomous decision-making remains limited by incomplete understanding of their internal mechanisms. As these mod…

Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing

2026-05-27 · Li Lei, Madalina Ciobanu, Qingqing Mao, Ritankar Das arxiv

LLMs increasingly require surgical model editing to enhance domain-specific capabilities without incurring the computational cost or catastrophic forgetting associated with full fine-tuning. Sparse Autoencoders (SAEs) ha…

Mathematical Reasoning

Latent Knowledge Scalpel: Precise and Massive Knowledge Editing for Large Language Models

2025-08-01 · Xin Liu, Qiyang Song, Shaowen Xu, Kerou Zhou 외 arxiv

Large Language Models (LLMs) often retain inaccurate or outdated information from pre-training, leading to incorrect predictions or biased outputs during inference. While existing model editing methods can address this c…

knowledge editing