Ian Berlot-Attwell

Ian Berlot-Attwell, Frank Rudzicz

In this work, we evaluate various existing dialogue relevance metrics, find strong dependency on the dataset, often with poor correlation with human scores of relevance, and propose modifications to reduce data requirements and domain sensitivity while improving correlation. Our proposed metric achieves state-of-the-art performance on the HUMOD dataset while reducing measured sensitivity to dataset by 37%-66%. We achieve this without fine-tuning a pretrained language model, and using only 3,750 unannotated human dialogues and a single negative example. Despite these limitations, we demonstrate competitive performance on four datasets from different domains. Our code, including our metric and experiments, is open sourced.

Ian Berlot-Attwell, Kumar Krishna Agrawal, A. Michael Carrell et al. · cmu, harvard

Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen and unseen combinations of object attributes. To test, we introduce a novel diagnostic dataset, CLEVR-HOPE. We find that the systematicity gap is not reduced by increasing the quantity of training data, but is reduced by increasing the diversity of training data. In particular, our experiments suggest that the more distinct attribute type combinations are seen during training, the more systematic we can expect the resulting model to be.

Panagiotis Lymperopoulos, Abhiramon Rajasekharan, Ian Berlot-Attwell et al.

Building world models is essential for planning in real-world domains such as businesses. Since such domains have rich semantics, we can leverage world knowledge to effectively model complex action effects and causal relationships from limited data. In this work, we propose CASSANDRA, a neurosymbolic world modeling approach that leverages an LLM as a knowledge prior to construct lightweight transition models for planning. CASSANDRA integrates two components: (1) LLM-synthesized code to model deterministic features, and (2) LLM-guided structure learning of a probabilistic graphical model to capture causal relationships among stochastic variables. We evaluate CASSANDRA in (i) a small-scale coffee-shop simulator and (ii) a complex theme park business simulator, where we demonstrate significant improvements in transition prediction and planning over baselines.

Ian Berlot-Attwell, Frank Rudzicz, Xujie Si

Advances in Large Language Models (LLMs) have spurred a wave of LLM library learning systems for mathematical reasoning. These systems aim to learn a reusable library of tools, such as formal Isabelle lemmas or Python programs that are tailored to a family of tasks. Many of these systems are inspired by the human structuring of knowledge into reusable and extendable concepts, but do current methods actually learn reusable libraries of tools? We study two library learning systems for mathematics which both reported increased accuracy: LEGO-Prover and TroVE. We find that function reuse is extremely infrequent on miniF2F and MATH. Our followup ablation experiments suggest that, rather than reuse, self-correction and self-consistency are the primary drivers of the observed performance gains. Our code and data are available at https://github.com/ikb-a/curious-case

Stéphane Aroca-Ouellette, Ian Berlot-Attwell, Panagiotis Lymperopoulos et al.

Despite rapid progress in artificial intelligence, current systems struggle with the interconnected challenges that define real-world decision making. Practical domains, such as business management, require optimizing an open-ended and multi-faceted objective, actively learning environment dynamics from sparse experience, planning over long horizons in stochastic settings, and reasoning over spatial information. Yet existing human--AI benchmarks isolate subsets of these capabilities, limiting our ability to assess holistic decision-making competence. We introduce Mini Amusement Parks (MAPs), an amusement-park simulator designed to evaluate an agent's ability to model its environment, anticipate long-term consequences under uncertainty, and strategically operate a complex business. We provide human baselines and a comprehensive evaluation of state-of-the-art LLM agents, finding that humans outperform these systems by 6.5x on easy mode and 9.8x on medium mode. Our analysis reveals persistent weaknesses in long-horizon optimization, sample-efficient learning, spatial reasoning, and world modelling. By unifying these challenges within a single environment, MAPs offers a new foundation for benchmarking agents capable of adaptable decision making. Code: https://github.com/Skyfall-Research/MAPs

Kaustubh D. Dhole, Varun Gangal, Sebastian Gehrmann et al.

Data augmentation is an important component in the robustness evaluation of models in natural language processing (NLP) and in enhancing the diversity of the data they are trained on. In this paper, we present NL-Augmenter, a new participatory Python-based natural language augmentation framework which supports the creation of both transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of natural language tasks. We demonstrate the efficacy of NL-Augmenter by using several of its transformations to analyze the robustness of popular natural language models. The infrastructure, datacards and robustness analysis results are available publicly on the NL-Augmenter repository (https://github.com/GEM-benchmark/NL-Augmenter).

John Chen, Ian Berlot-Attwell, Safwan Hossain et al.

Clinical machine learning is increasingly multimodal, collected in both structured tabular formats and unstructured forms such as freetext. We propose a novel task of exploring fairness on a multimodal clinical dataset, adopting equalized odds for the downstream medical prediction tasks. To this end, we investigate a modality-agnostic fairness algorithm - equalized odds post processing - and compare it to a text-specific fairness algorithm: debiased clinical word embeddings. Despite the fact that debiased word embeddings do not explicitly address equalized odds of protected groups, we show that a text-specific approach to fairness may simultaneously achieve a good balance of performance and classical notions of fairness. We hope that our paper inspires future contributions at the critical intersection of clinical NLP and fairness. The full source code is available here: https://github.com/johntiger1/multimodal_fairness

Ian Berlot-Attwell, Frank Rudzicz, Xujie Si

Recent advancements in the coding, reasoning, and tool-using abilities of LLMs have spurred interest in library learning (i.e., online learning through the creation, storage, and retrieval of reusable and composable functions, knowledge, checklists, or lemmas). Such systems often promise improved task performance through the automatic creation of broadly applicable tools, as well as superior computational performance through the caching of reasoning (i.e., the storage of generated tools). However, we find strong reason to be skeptical. We perform a deep dive into one such system, LEGO-Prover, which purports to learn reusable lemmas for mathematical reasoning. We find no evidence of the direct reuse of learned lemmas, and find evidence against the soft reuse of learned lemmas (i.e., reuse by modifying relevant examples). Crucially, we find that LEGO-Prover does not in fact improve over the simple baseline of prompting the model - the improvements in task accuracy vanish once computational cost is accounted for. Our findings suggest that serious misconceptions exist as to the effectiveness of these techniques, that a serious re-examination of the state of LLM-based library learning is required, and that we require much stronger standards for evaluation including behavioural analysis and ensuring that an equal computational budget is used for baselines.

Tobias Sesterhenn, Ian Berlot-Attwell, Janis Zenkner et al.

Reusing established theorems and formulas is central to mathematical problem solving, serving as essential building blocks for tackling increasingly complex challenges. Recent work, TroVE, argues that code-generating Large Language Models (LLMs) can benefit similarly on the MATH benchmark by inducing and reusing higher-level toolboxes. By allocating computational budget across an ensemble of three modes -- directly generating code, creating tools, and reusing tools -- TroVE claims to outperform a PRIMITIVE baseline that only performs direct generation. However, recent analysis (Berlot-Attwell et al., 2024) casts doubt on these gains, noting that the tools created are often trivial or rarely reused, suggesting that improvements may stem from self-consistency or self-correction. In this work, we re-evaluate TroVE on MATH, analyze the impact of each of its modes, and show that its benefit does not come from these mechanisms, but simply from a higher computational budget spent for TroVE compared to PRIMITIVE. To this end, we also perform a small correction in the original implementation of TroVE's selection mechanism, boosting TroVE's performance on MATH by 3\% in accuracy. After matching for compute, the benefit of TroVE reduces to a marginal improvement of 1\%, suggesting that this toolbox approach does not provide a significant benefit on MATH.

Ian Berlot-Attwell

An ultimate goal of the AI and ML fields is artificial general intelligence (AGI); although such systems remain science fiction, various models exhibit aspects of AGI. In this work, we look at neuro-symbolic (NS)approaches to visual question answering (VQA) from the perspective of AGI desiderata. We see how well these systems meet these desiderata, and how the desiderata often pull the scientist in opposing directions. It is my hope that through this work we can temper model evaluation on benchmarks with a discussion of the properties of these systems and their potential for future extension.

Ian Berlot-Attwell, Frank Rudzicz

Automatically evaluating text-based, non-task-oriented dialogue systems (i.e., `chatbots') remains an open problem. Previous approaches have suffered challenges ranging from poor correlation with human judgment to poor generalization and have often required a gold standard reference for comparison or human-annotated data. Extending existing evaluation methods, we propose that a metric based on linguistic features may be able to maintain good correlation with human judgment and be interpretable, without requiring a gold-standard reference or human-annotated data. To support this proposition, we measure and analyze various linguistic features on dialogues produced by multiple dialogue models. We find that the features' behaviour is consistent with the known properties of the models tested, and is similar across domains. We also demonstrate that this approach exhibits promising properties such as zero-shot generalization to new domains on the related task of evaluating response relevance.