Reward models have become a staple in modern NLP, serving as not only a scalable text evaluator, but also an indispensable component in many alignment recipes and inference-time algorithms. However, while recent reward models increase performance on standard benchmarks, this may partly be due to overfitting effects, which would confound an understanding of their true capability. In this work, we scrutinize the robustness of reward models and the extent of such overfitting. We build **reWordBench**, which systematically transforms reward model inputs in meaning- or ranking-preserving ways. We show that state-of-the-art reward models suffer from substantial performance degradation even with minor input transformations, sometimes dropping to significantly below-random accuracy, suggesting brittleness. To improve reward model robustness, we propose to explicitly train them to assign similar scores to paraphrases, and find that this approach also improves robustness to other distinct kinds of transformations. For example, our robust reward model reduces such degradation by roughly half for the Chat Hard subset in RewardBench. Furthermore, when used in alignment, our robust reward models demonstrate better utility and lead to higher-quality outputs, winning in up to 59% of instances against a standardly trained RM.

We introduce SelfCite, a novel self-supervised approach that aligns LLMs to generate high-quality, fine-grained, sentence-level citations for the statements in their generated responses. Instead of only relying on costly and labor-intensive annotations, SelfCite leverages a reward signal provided by the LLM itself through context ablation: If a citation is necessary, removing the cited text from the context should prevent the same response; if sufficient, retaining the cited text alone should preserve the same response. This reward can guide the inference-time best-of-N sampling strategy to improve citation quality significantly, as well as be used in preference optimization to directly fine-tune the models for generating better citations. The effectiveness of SelfCite is demonstrated by increasing citation F1 up to 5.3 points on the LongBench-Cite benchmark across five long-form question answering tasks.

Modern language models can process inputs across diverse languages and modalities. We hypothesize that models acquire this capability through learning a shared representation space across heterogeneous data types (e.g., different languages and modalities), which places semantically similar inputs near one another, even if they are from different modalities/languages. We term this the semantic hub hypothesis, following the hub-and-spoke model from neuroscience (Patterson et al., 2007) which posits that semantic knowledge in the human brain is organized through a transmodal semantic "hub" which integrates information from various modality-specific "spokes" regions. We first show that model representations for semantically equivalent inputs in different languages are similar in the intermediate layers, and that this space can be interpreted using the model's dominant pretraining language via the logit lens. This tendency extends to other data types, including arithmetic expressions, code, and visual/audio inputs. Interventions in the shared representation space in one data type also predictably affect model outputs in other data types, suggesting that this shared representations space is not simply a vestigial byproduct of large-scale training on broad data, but something that is actively utilized by the model during input processing. For every other layer, we show the closest output token to the hidden state based on the logit lens. Llama-3's hidden states are often closest to English tokens when processing Chinese texts, arithmetic expressions, and code, in a semantically corresponding way. LLaVA, a vision-language model, and SALMONN, an audio-language model, have similar behavior when processing images/audio. As shown for the arithmetic expression example, models can be intervened cross-lingually or cross-modally, such as using English even though the input is non-English, and be steered towards corresponding effects. Boldface is only for emphasis. How do LMs process these distinct data types with a single set of parameters? One strategy might be to learn specialized subspaces for each data type that are only employed when processing it. In many cases, however, data types that are surface-distinct share underlying semantic concepts.

Vision language models (VLMs) have demonstrated impressive performance across a wide range of downstream tasks. However, their proficiency in spatial reasoning remains limited, despite its crucial role in tasks involving navigation and interaction with physical environments. Specifically, most of these tasks rely on the core spatial reasoning capabilities in two-dimensional (2D) environments, and our evaluation reveals that state-of-the-art VLMs frequently generate implausible and incorrect responses to composite spatial reasoning problems, including simple pathfinding tasks that humans can solve effortlessly at a glance. To address this, we explore an effective approach to enhance 2D spatial reasoning within VLMs by training the model solely on basic spatial capabilities. We begin by disentangling the key components of 2D spatial reasoning: direction comprehension, distance estimation, and localization. Our central hypothesis is that mastering these basic spatial capabilities can significantly enhance a model's performance on composite spatial tasks requiring advanced spatial understanding and combinatorial problem-solving, with generalized improvements in visual-spatial tasks. To investigate this hypothesis, we introduce Sparkle, a framework that fine-tunes VLMs on these three basic spatial capabilities by synthetic data generation and targeted supervision to form an instruction dataset for each capability. Our experiments demonstrate that VLMs fine-tuned with Sparkle achieve significant performance gains, not only in the basic tasks themselves but also in generalizing to composite and out-of-distribution spatial reasoning tasks. These findings underscore the effectiveness of mastering basic spatial capabilities in enhancing composite spatial problem-solving, offering insights into systematic strategies for improving VLMs' spatial reasoning capabilities.

Aligning language models (LMs) based on human-annotated preference data is a crucial step in obtaining practical and performant LM-based systems. However, multilingual human preference data are difficult to obtain at scale, making it challenging to extend this framework to diverse languages. In this work, we evaluate a simple approach for zero-shot cross-lingual alignment, where a reward model is trained on preference data in one source language and directly applied to other target languages. On summarization and open-ended dialog generation, we show that this method is consistently successful under comprehensive evaluation settings, including human evaluation: cross-lingually aligned models are preferred by humans over unaligned models on up to >70% of evaluation instances. We moreover find that a different-language reward model sometimes yields better aligned models than a same-language reward model. We also identify best practices when there is no language-specific data for even supervised finetuning, another component in alignment.

Ambiguity is an critical component of language that allows for more effective communication between speakers, but is often ignored in NLP. Recent work suggests that NLP systems may struggle to grasp certain elements of human language understanding because they may not handle ambiguities at the level that humans naturally do in communication. Additionally, different types of ambiguity may serve different purposes and require different approaches for resolution, and we aim to investigate how language models' abilities vary across types. We propose a taxonomy of ambiguity types as seen in English to facilitate NLP analysis. Our taxonomy can help make meaningful splits in language ambiguity data, allowing for more fine-grained assessments of both datasets and model performance.

Do LMs infer the semantics of text from co-occurrence patterns in their training data? Merrill et al. (2022) argue that, in theory, probabilities predicted by an optimal LM encode semantic information about entailment relations, but it is unclear whether neural LMs trained on corpora learn entailment in this way because of strong idealizing assumptions made by Merrill et al. In this work, we investigate whether their theory can be used to decode entailment judgments from neural LMs. We find that a test similar to theirs can decode entailment relations between natural sentences, well above random chance, though not perfectly, across many datasets and LMs. This suggests LMs implicitly model aspects of semantics to predict semantic effects on sentence co-occurrence patterns. However, we find the test that predicts entailment in practice works in the opposite direction to the theoretical test. We thus revisit the assumptions underlying the original test, finding its derivation did not adequately account for redundancy in human-written text. We argue that correctly accounting for redundancy related to explanations might derive the observed flipped test and, more generally, improve linguistic theories of human speakers.

Thin film deposition is an essential step in the semiconductor process. During preparation or loading, the substrate is exposed to the air unavoidably, which has motivated studies of the process control to remove the surface oxide before thin film deposition. Optimizing the deoxidation process in molecular beam epitaxy (MBE) for a random substrate is a multidimensional challenge and sometimes controversial. Due to variations in semiconductor materials and growth processes, the determination of substrate deoxidation temperature is highly dependent on the grower's expertise; the same substrate may yield inconsistent results when evaluated by different growers. Here, we employ a machine learning (ML) hybrid convolution and vision transformer (CNN-ViT) model. This model utilizes reflection high-energy electron diffraction (RHEED) video as input to determine the deoxidation status of the substrate as output, enabling automated substrate deoxidation under a controlled architecture. This also extends to the successful application of deoxidation processes on other substrates. Furthermore, we showcase the potential of models trained on data from a single MBE equipment to achieve high-accuracy deployment on other equipment. In contrast to traditional methods, our approach holds exceptional practical value. It standardizes deoxidation temperatures across various equipment and substrate materials, advancing the standardization research process in semiconductor preparation, a significant milestone in thin film growth technology. The concepts and methods demonstrated in this work are anticipated to revolutionize semiconductor manufacturing in optoelectronics and microelectronics industries by applying them to diverse material growth processes.

Ambiguity is an intrinsic feature of natural language. Managing ambiguity is a key part of human language understanding, allowing us to anticipate misunderstanding as communicators and revise our interpretations as listeners. As language models (LMs) are increasingly employed as dialogue interfaces and writing aids, handling ambiguous language is critical to their success. We characterize ambiguity in a sentence by its effect on entailment relations with another sentence, and collect AmbiEnt, a linguist-annotated benchmark of 1,645 examples with diverse kinds of ambiguity. We design a suite of tests based on AmbiEnt, presenting the first evaluation of pretrained LMs to recognize ambiguity and disentangle possible meanings. We find that the task remains extremely challenging, including for GPT-4, whose generated disambiguations are considered correct only 32% of the time in human evaluation, compared to 90% for disambiguations in our dataset. Finally, to illustrate the value of ambiguity-sensitive tools, we show that a multilabel NLI model can flag political claims in the wild that are misleading due to ambiguity. We encourage the field to rediscover the importance of ambiguity for NLP.

Self-assembled InAs/GaAs quantum dots (QDs) have properties highly valuable for developing various optoelectronic devices such as QD lasers and single photon sources. The applications strongly rely on the density and quality of these dots, which has motivated studies of the growth process control to realize high-quality epi-wafers and devices. Establishing the process parameters in molecular beam epitaxy (MBE) for a specific density of QDs is a multidimensional optimization challenge, usually addressed through time-consuming and iterative trial-and-error. Here, we report a real-time feedback control method to realize the growth of QDs with arbitrary density, which is fully automated and intelligent. We developed a machine learning (ML) model named 3D ResNet 50 trained using reflection high-energy electron diffraction (RHEED) videos as input instead of static images and providing real-time feedback on surface morphologies for process control. As a result, we demonstrated that ML from previous growth could predict the post-growth density of QDs, by successfully tuning the QD densities in near-real time from 1.5E10 cm-2 down to 3.8E8 cm-2 or up to 1.4E11 cm-2. Compared to traditional methods, our approach, with in situ tuning capabilities and excellent reliability, can dramatically expedite the material optimization process and improve the reproducibility of MBE, constituting significant progress for thin film growth techniques. The concepts and methodologies proved feasible in this work are promising to be applied to a variety of material growth processes, which will revolutionize semiconductor manufacturing for optoelectronic and microelectronic industries.