565 citations · 2.5k across the 94 of their papers we have counts for
108 papers
LIFT: Scoped Constraints for Evaluating and Training Instruction Following
Eran Goldstein, Yuval Belfer, Yonatan Belinkov +1
Common instruction following (IF) benchmarks, such as IFEval, evaluate the ability of language models (LMs) to follow simple constraints that can easily be verified programmaticall…
Feeding BabyLMs Macaroni: Code-Switching Curricula Cause Cross-Lingual Convergence
Dries Rooryck, Alex Cai, Yonatan Belinkov +2
Children in multilingual communities often code-switch, using multiple languages in a single utterance. Can we induce cross-lingual alignment in language models by training on code…
Decomposing Query-Key Feature Interactions Using Contrastive Covariances
Andrew Lee, Yonatan Belinkov, Fernanda Viégas +1
Despite the central role of attention heads in Transformers, we lack tools to understand why a model attends to a particular token. To address this, we study the query-key (QK) spa…
Investigating the Development of Task-Oriented Communication in Vision-Language Models
Boaz Carmeli, Orr Paradise, Shafi Goldwasser +2
We investigate whether \emph{LLM-based agents} can develop task-oriented communication protocols that differ from standard natural language in collaborative reasoning tasks. Our fo…
CtD: Composition through Decomposition in Emergent Communication
Boaz Carmeli, Ron Meir, Yonatan Belinkov
Compositionality is a cognitive mechanism that allows humans to systematically combine known concepts in novel ways. This study demonstrates how artificial neural agents acquire an…
Will it Merge? On The Causes of Model Mergeability
Adir Rahamim, Asaf Yehudai, Boaz Carmeli +3
Model merging has emerged as a promising technique for combining multiple fine-tuned models into a single multitask model without retraining. However, the factors that determine wh…
Structured RAG for Answering Aggregative Questions
Omri Koshorek, Niv Granot, Aviv Alloni +6
Retrieval-Augmented Generation (RAG) has become the dominant approach for answering questions over large corpora. However, current datasets and methods are highly focused on cases…
Planted in Pretraining, Swayed by Finetuning: A Case Study on the Origins of Cognitive Biases in LLMs
Itay Itzhak, Yonatan Belinkov, Gabriel Stanovsky
Large language models (LLMs) exhibit cognitive biases -- systematic tendencies of irrational decision-making, similar to those seen in humans. Prior work has found that these biase…
Reverse-Engineering the Retrieval Process in GenIR Models
Anja Reusch, Yonatan Belinkov
Generative Information Retrieval (GenIR) is a novel paradigm in which a transformer encoder-decoder model predicts document rankings based on a query in an end-to-end fashion. Thes…
Padding Tone: A Mechanistic Analysis of Padding Tokens in T2I Models
Michael Toker, Ido Galil, Hadas Orgad +4
Text-to-image (T2I) diffusion models rely on encoded prompts to guide the image generation process. Typically, these prompts are extended to a fixed length by adding padding tokens…
Back Attention: Understanding and Enhancing Multi-Hop Reasoning in Large Language Models
Zeping Yu, Yonatan Belinkov, Sophia Ananiadou
We investigate how large language models perform latent multi-hop reasoning in prompts like "Wolfgang Amadeus Mozart's mother's spouse is". To analyze this process, we introduce lo…
Unsupervised Translation of Emergent Communication
Ido Levy, Orr Paradise, Boaz Carmeli +3
Emergent Communication (EC) provides a unique window into the language systems that emerge autonomously when agents are trained to jointly achieve shared goals. However, it is diff…
Position-aware Automatic Circuit Discovery
Tal Haklay, Hadas Orgad, David Bau +2
A widely used strategy to discover and understand language model mechanisms is circuit analysis. A circuit is a minimal subgraph of a model's computation graph that executes a spec…
Trust Me, I'm Wrong: LLMs Hallucinate with Certainty Despite Knowing the Answer
Adi Simhi, Itay Itzhak, Fazl Barez +2
Prior work on large language model (LLM) hallucinations has associated them with model uncertainty or inaccurate knowledge. In this work, we define and investigate a distinct type…
Semantics and Spatiality of Emergent Communication
Rotem Ben Zion, Boaz Carmeli, Orr Paradise +1
When artificial agents are jointly trained to perform collaborative tasks using a communication channel, they develop opaque goal-oriented communication protocols. Good task perfor…
Growing a Tail: Increasing Output Diversity in Large Language Models
Michal Shur-Ofry, Bar Horowitz-Amsalem, Adir Rahamim +1
How diverse are the outputs of large language models when diversity is desired? We examine the diversity of responses of several language models to questions with multiple possible…
Context-aware Prompt Tuning: Advancing In-Context Learning with Adversarial Methods
Tsachi Blau, Moshe Kimhi, Yonatan Belinkov +2
Fine-tuning Large Language Models (LLMs) typically involves updating at least a few billions of parameters. A more parameter-efficient approach is Prompt Tuning (PT), which updates…
Distinguishing Ignorance from Error in LLM Hallucinations
Adi Simhi, Jonathan Herzig, Idan Szpektor +1
Large language models (LLMs) are susceptible to hallucinations -- factually incorrect outputs -- leading to a large body of work on detecting and mitigating such cases. We argue th…
Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics
Yaniv Nikankin, Anja Reusch, Aaron Mueller +1
Do large language models (LLMs) solve reasoning tasks by learning robust generalizable algorithms, or do they memorize training data? To investigate this question, we use arithmeti…
LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations
Hadas Orgad, Michael Toker, Zorik Gekhman +4
Large language models (LLMs) often produce errors, including factual inaccuracies, biases, and reasoning failures, collectively referred to as "hallucinations". Recent studies have…
Fast Forwarding Low-Rank Training
Adir Rahamim, Naomi Saphra, Sara Kangaslahti +1
Parameter efficient finetuning methods like low-rank adaptation (LoRA) aim to reduce the computational costs of finetuning pretrained Language Models (LMs). Enabled by these low-ra…
Jamba-1.5: Hybrid Transformer-Mamba Models at Scale
Jamba Team, Barak Lenz, Alan Arazi +58
We present Jamba-1.5, new instruction-tuned large language models based on our Jamba architecture. Jamba is a hybrid Transformer-Mamba mixture of experts architecture, providing hi…
Confidence Regulation Neurons in Language Models
Alessandro Stolfo, Ben Wu, Wes Gurnee +4
Despite their widespread use, the mechanisms by which large language models (LLMs) represent and regulate uncertainty in next-token predictions remain largely unexplored. This stud…
REVS: Unlearning Sensitive Information in Language Models via Rank Editing in the Vocabulary Space
Tomer Ashuach, Martin Tutek, Yonatan Belinkov
Language models (LMs) risk inadvertently memorizing and divulging sensitive or personally identifiable information (PII) seen in training data, causing privacy concerns. Current ap…