Imagen de Portada

Causality $\neq$ Invariance: Function and Concept Vectors in LLMs

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Opiełka, Gustaw, Rosenbusch, Hannes, Stevenson, Claire E.
Formato:	Preprint
Publicado:	2026
Materias:	Computation and Language Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Analogical Reasoning Inside Large Language Models: Concept Vectors and the Limits of Abstraction
por: Opiełka, Gustaw, et al.
Publicado: (2025)

Do Large Language Models Solve ARC Visual Analogies Like People Do?
por: Opiełka, Gustaw, et al.
Publicado: (2024)

Beyond Single Concept Vector: Modeling Concept Subspace in LLMs with Gaussian Distribution
por: Zhao, Haiyan, et al.
Publicado: (2024)

What makes for an enjoyable protagonist? An analysis of character warmth and competence
por: Rosenbusch, Hannes
Publicado: (2026)

Transformer See, Transformer Do: Copying as an Intermediate Step in Learning Analogical Reasoning
por: Hellwig, Philipp, et al.
Publicado: (2026)

Vector Quantized Latent Concepts: A Scalable Alternative to Clustering-Based Concept Discovery
por: Yu, Xuemin, et al.
Publicado: (2026)

LLM_annotate: A Python package for annotating and analyzing fiction characters
por: Rosenbusch, Hannes
Publicado: (2025)

Saliency Suppressed, Semantics Surfaced: Visual Transformations in Neural Networks and the Brain
por: Opiełka, Gustaw, et al.
Publicado: (2024)

Function Vectors in Large Language Models
por: Todd, Eric, et al.
Publicado: (2023)

On Lexical Invariance on Multisets and Graphs
por: Zhang, Muhan
Publicado: (2024)

Failure Modes of LLMs for Causal Reasoning on Narratives
por: Yamin, Khurram, et al.
Publicado: (2024)

CRVQ: Channel-Relaxed Vector Quantization for Extreme Compression of LLMs
por: Xu, Yuzhuang, et al.
Publicado: (2024)

Mathematical Reasoning via Intervention-Based Time-Series Causal Discovery Using LLMs as Concept Mastery Simulators
por: Okita, Tsuyoshi
Publicado: (2026)

Trained on Tokens, Calibrated on Concepts: The Emergence of Semantic Calibration in LLMs
por: Nakkiran, Preetum, et al.
Publicado: (2025)

Can LLMs Learn New Concepts Incrementally without Forgetting?
por: Zheng, Junhao, et al.
Publicado: (2024)

Relational Knowledge Distillation Using Fine-tuned Function Vectors
por: Kang, Andrea, et al.
Publicado: (2026)

Refusal in LLMs is an Affine Function
por: Marshall, Thomas, et al.
Publicado: (2024)

Cyclic Ablation: Testing Concept Localization against Functional Regeneration in AI
por: Kapelko, Eduard
Publicado: (2025)

Causally Testing Gender Bias in LLMs: A Case Study on Occupational Bias
por: Chen, Yuen, et al.
Publicado: (2022)

NEAT: Concept driven Neuron Attribution in LLMs
por: Kavuri, Vivek Hruday, et al.
Publicado: (2025)

Self-Supervised Learning for Neural Topic Models with Variance-Invariance-Covariance Regularization
por: Xu, Weiran, et al.
Publicado: (2025)

Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs
por: Siddique, Zara, et al.
Publicado: (2025)

Can Post-Training Transform LLMs into Causal Reasoners?
por: Chen, Junqi, et al.
Publicado: (2026)

CauScientist: Teaching LLMs to Respect Data for Causal Discovery
por: Peng, Bo, et al.
Publicado: (2026)

Are the Values of LLMs Structurally Aligned with Humans? A Causal Perspective
por: Kang, Yipeng, et al.
Publicado: (2024)

Are some books better than others?
por: Rosenbusch, Hannes, et al.
Publicado: (2025)

MedConceptsQA: Open Source Medical Concepts QA Benchmark
por: Shoham, Ofir Ben, et al.
Publicado: (2024)

Kernelized Concept Erasure
por: Ravfogel, Shauli, et al.
Publicado: (2022)

Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMs
por: Han, Pengrui, et al.
Publicado: (2026)

Which books do I like?
por: Rosenbusch, Hannes, et al.
Publicado: (2025)

CausalEvolve: Towards Open-Ended Discovery with Causal Scratchpad
por: Chen, Yongqiang, et al.
Publicado: (2026)

Causality $\neq$ Decodability, and Vice Versa: Lessons from Interpreting Counting ViTs
por: Huang, Lianghuan, et al.
Publicado: (2025)

Towards Compositionality in Concept Learning
por: Stein, Adam, et al.
Publicado: (2024)

Linear Adversarial Concept Erasure
por: Ravfogel, Shauli, et al.
Publicado: (2022)

LLM Pretraining with Continuous Concepts
por: Tack, Jihoon, et al.
Publicado: (2025)

Predicting Where Steering Vectors Succeed
por: Billa, Jayadev
Publicado: (2026)

Concurrency without Model Changes: Future-based Asynchronous Function Calling for LLMs
por: Feng, Guangyu, et al.
Publicado: (2026)

One Example Shown, Many Concepts Known! Counterexample-Driven Conceptual Reasoning in Mathematical LLMs
por: Li, Yinghui, et al.
Publicado: (2025)

Concept Bottleneck Large Language Models
por: Sun, Chung-En, et al.
Publicado: (2024)

Adaptive Task Vectors for Large Language Models
por: Kang, Joonseong, et al.
Publicado: (2025)