Robert West

AAAI Conference 2026 Conference Paper

Interactive Evaluation of Large Language Models for Multi-Requirement Software Engineering Tasks

• Dimitrios Rontogiannis
• Maxime Peyrard
• Nicolas Baldwin
• Martin Josifoski
• Robert West
• Dimitrios Gunopulos

Standard single-turn, static benchmarks fall short in evaluating the nuanced capabilities of Large Language Models (LLMs) on complex tasks such as software engineering. In this work, we propose a novel interactive evaluation framework that assesses LLMs on multi-requirement programming tasks through structured, feedback-driven dialogue. Each task is modeled as a requirement dependency graph, and an "interviewer" LLM, aware of the ground-truth solution, provides minimal, targeted hints to an "interviewee" model to help correct errors and fulfill target constraints. This dynamic protocol enables fine-grained diagnostic insights into model behavior, uncovering strengths and systematic weaknesses that static benchmarks fail to measure. We build on DevAI, a benchmark of 55 curated programming tasks, by adding ground-truth solutions and evaluating the relevance and utility of interviewer hints through expert annotation. Our results highlight the importance of dynamic evaluation in advancing the development of collaborative code-generating agents.

NeurIPS Conference 2025 Conference Paper

One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models

• Viacheslav Surkov
• Chris Wendler
• Antonio Mari
• Mikhail Terekhov
• Justin Deschenaux
• Robert West
• Caglar Gulcehre
• David Bau

For large language models (LLMs), sparse autoencoders (SAEs) have been shown to decompose intermediate representations that often are not interpretable directly into sparse sums of interpretable features, facilitating better control and subsequent analysis. However, similar analyses and approaches have been lacking for text-to-image models. We investigate the possibility of using SAEs to learn interpretable features for SDXL Turbo, a few-step text-to-image diffusion model. To this end, we train SAEs on the updates performed by transformer blocks within SDXL Turbo's denoising U-net in its 1-step setting. Interestingly, we find that they generalize to 4-step SDXL Turbo and even to the multi-step SDXL base model (i. e. , a different model) without additional training. In addition, we show that their learned features are interpretable, causally influence the generation process, and reveal specialization among the blocks. We do so by creating RIEBench, a representation-based image editing benchmark, for editing images while they are generated by turning on and off individual SAE features. This allows us to track which transformer blocks' features are the most impactful depending on the edit category. Our work is the first investigation of SAEs for interpretability in text-to-image diffusion models and our results establish SAEs as a promising approach for understanding and manipulating the internal mechanisms of text-to-image models.

NeurIPS Conference 2025 Conference Paper

zip2zip: Inference-Time Adaptive Tokenization via Online Compression

• Saibo Geng
• Nathan Ranchin
• Yunzhen Yao
• Maxime Peyrard
• Chris Wendler
• Michael Gastpar
• Robert West

Tokenization efficiency plays a critical role in the performance and cost of large language models (LLMs), yet most models rely on static tokenizers optimized on general-purpose corpora. These tokenizers’ fixed vocabularies often fail to adapt to domain- or language-specific inputs, leading to longer token sequences and higher computational costs. We introduce zip2zip, a novel method for achieving context-adaptive tokenization in LLMs at inference time. Leveraging an online data compression algorithm (Lempel–Ziv–Welch), zip2zip dynamically expands its active vocabulary at inference time by continuously replacing fragmented token sequences with more compact hypertokens, which it can immediately output during generation. In doing so, the model refines its internal tokenization scheme to match the token distribution of the current context, reducing redundancy and improving representational efficiency. zip2zip consists of three key components: (1) a tokenizer based on Lempel–Ziv–Welch compression that incrementally merges co-occurring tokens into reusable hypertokens on the fly; (2) a dynamic embedding (and unembedding) layer that computes embeddings for newly formed hypertokens at runtime; and (3) a variant of autoregressive language modeling that pretrains the model to handle hypertokenized, compressed text sequences as inputs and outputs. We show that an existing LLM can be uptrained for zip2zip in 10 GPU-hours via parameter-efficient finetuning. The resulting LLM performs test-time adaptation, learning to use hypertokens in unseen contexts and reducing input and output tokens by 15–40%. Code and models are released at https: //github. com/epfl-dlab/zip2zip.

NeSy Conference 2022 Conference Paper

ESC-Rules: Explainable, Semantically Constrained Rule Sets

• Martin Glauer
• Robert West
• Susan Michie
• Janna Hastings

We describe a novel approach to explainable prediction of a continuous variable based on learning fuzzy weighted rules. Our model trains a set of weighted rules to maximise prediction accuracy and minimise an ontology-based ’semantic loss’ function including user-specified constraints on the rules that should be learned in order to maximise the explainability of the resulting rule set from a user perspective. This system fuses quantitative sub-symbolic learning with symbolic learning and constraints based on domain knowledge. We illustrate our system on a case study in predicting the outcomes of behavioural interventions for smoking cessation, and show that it outperforms other interpretable approaches, achieving performance close to that of a deep learning model, while offering transparent explainability that is an essential requirement for decision-makers in the health domain.

IJCAI Conference 2021 Conference Paper

A Ladder of Causal Distances

• Maxime Peyrard
• Robert West

Causal discovery, the task of automatically constructing a causal model from data, is of major significance across the sciences. Evaluating the performance of causal discovery algorithms should ideally involve comparing the inferred models to ground-truth models available for benchmark datasets, which in turn requires a notion of distance between causal models. While such distances have been proposed previously, they are limited by focusing on graphical properties of the causal models being compared. Here, we overcome this limitation by defining distances derived from the causal distributions induced by the models, rather than exclusively from their graphical structure. Pearl and Mackenzie [2018] have arranged the properties of causal models in a hierarchy called the ``ladder of causation'' spanning three rungs: observational, interventional, and counterfactual. Following this organization, we introduce a hierarchy of three distances, one for each rung of the ladder. Our definitions are intuitively appealing as well as efficient to compute approximately. We put our causal distances to use by benchmarking standard causal discovery systems on both synthetic and real-world datasets for which ground-truth causal models are available.

IJCAI Conference 2021 Conference Paper

Laughing Heads: Can Transformers Detect What Makes a Sentence Funny?

• Maxime Peyrard
• Beatriz Borges
• Kristina Gligorić
• Robert West

The automatic detection of humor poses a grand challenge for natural language processing. Transformer-based systems have recently achieved remarkable results on this task, but they usually (1) were evaluated in setups where serious vs humorous texts came from entirely different sources, and (2) focused on benchmarking performance without providing insights into how the models work. We make progress in both respects by training and analyzing transformer-based humor recognition models on a recently introduced dataset consisting of minimal pairs of aligned sentences, one serious, the other humorous. We find that, although our aligned dataset is much harder than previous datasets, transformer-based models recognize the humorous sentence in an aligned pair with high accuracy (78\%). In a careful error analysis, we characterize easy vs hard instances. Finally, by analyzing attention weights, we obtain important insights into the mechanisms by which transformers recognize humor. Most remarkably, we find clear evidence that one single attention head learns to recognize the words that make a test sentence humorous, even without access to this information at training time.

AAAI Conference 2019 Conference Paper

Reverse-Engineering Satire, or “Paper on Computational Humor Accepted despite Making Serious Advances”

• Robert West
• Eric Horvitz

Humor is an essential human trait. Efforts to understand humor have called out links between humor and the foundations of cognition, as well as the importance of humor in social engagement. As such, it is a promising and important subject of study, with relevance for artificial intelligence and human– computer interaction. Previous computational work on humor has mostly operated at a coarse level of granularity, e. g. , predicting whether an entire sentence, paragraph, document, etc. , is humorous. As a step toward deep understanding of humor, we seek fine-grained models of attributes that make a given text humorous. Starting from the observation that satirical news headlines tend to resemble serious news headlines, we build and analyze a corpus of satirical headlines paired with nearly identical but serious headlines. The corpus is constructed via Unfun. me, an online game that incentivizes players to make minimal edits to satirical headlines with the goal of making other players believe the results are serious headlines. The edit operations used to successfully remove humor pinpoint the words and concepts that play a key role in making the original, satirical headline funny. Our analysis reveals that the humor tends to reside toward the end of headlines, and primarily in noun phrases, and that most satirical headlines follow a certain logical pattern, which we term false analogy. Overall, this paper deepens our understanding of the syntactic and semantic structure of satirical news headlines and provides insights for building humor-producing systems.

YNIMG Journal 2016 Journal Article

Electrophysiological correlates of error initiation and response correction

• Paul D. Kieffaber
• Jamie Hershaw
• Jonathan Sredl
• Robert West

NeurIPS Conference 2013 Conference Paper

Compressive Feature Learning

• Hristo Paskov
• Robert West
• John Mitchell
• Trevor Hastie

This paper addresses the problem of unsupervised feature learning for text data. Our method is grounded in the principle of minimum description length and uses a dictionary-based compression scheme to extract a succinct feature set. Specifically, our method finds a set of word $k$-grams that minimizes the cost of reconstructing the text losslessly. We formulate document compression as a binary optimization task and show how to solve it approximately via a sequence of reweighted linear programs that are efficient to solve and parallelizable. As our method is unsupervised, features may be extracted once and subsequently used in a variety of tasks. We demonstrate the performance of these features over a range of scenarios including unsupervised exploratory analysis and supervised text categorization. Our compressed feature space is two orders of magnitude smaller than the full $k$-gram space and matches the text categorization accuracy achieved in the full feature space. This dimensionality reduction not only results in faster training times, but it can also help elucidate structure in unsupervised learning tasks and reduce the amount of training data necessary for supervised learning.

IJCAI Conference 2009 Conference Paper

• Robert West
• Joelle Pineau
• Doina Precup

Computing the semantic distance between realworld concepts is crucial for many intelligent applications. We present a novel method that leverages data from ‘Wikispeedia’, an online game played on Wikipedia; players have to reach an article from another, unrelated article, only by clicking links in the articles encountered. In order to automatically infer semantic distances between everyday concepts, our method effectively extracts the common sense displayed by humans during play, and is thus more desirable, from a cognitive point of view, than purely corpus-based methods. We show that our method significantly outperforms Latent Semantic Analysis in a psychometric evaluation of the quality of learned semantic distances.