We examine whether large language models (LLMs) exhibit race- and gender-based name discrimination in hiring decisions, similar to classic findings in the social sciences (Bertrand and Mullainathan, 2004). We design a series of templatic prompts to LLMs to write an email to a named job applicant informing them of a hiring decision. By manipulating the applicant's first name, we measure the effect of perceived race, ethnicity, and gender on the probability that the LLM generates an acceptance or rejection email. We find that the hiring decisions of LLMs in many settings are more likely to favor White applicants over Hispanic applicants. In aggregate, the groups with the highest and lowest acceptance rates respectively are masculine White names and masculine Hispanic names. However, the comparative acceptance rates by group vary under different templatic settings, suggesting that LLMs' race- and gender-sensitivity may be idiosyncratic and prompt-sensitive.

Large language models (LLMs) have demonstrated impressive task-solving capabilities, achieved through either prompting techniques or system designs. However, concerns have arisen regarding their proficiency in planning tasks, as they often struggle to generate valid plans. This paper investigates the impact of fine-tuning on LLMs' planning capabilities. Our findings indicate that LLMs can achieve good performance in planning through substantial (thousands of specific examples) fine-tuning. However, fine-tuning is associated with significant economic and computational costs. To address this challenge, we propose the Maximum Diversity Fine-Tuning (MDFT) strategy to improve the sample efficiency of fine-tuning in the planning domain. Specifically, our algorithm, referred to as MDFT-g, encodes the planning task instances with their graph representations and selects a subset of samples in the vector space that maximizes data diversity. We empirically demonstrate that MDFT-g consistently outperforms existing baselines at various scales across multiple benchmark domains.

In this work, we employ two AMR-enhanced semantic representations for ICL on RE: one that explores the AMR structure generated for a sentence at the subgraph level (shortest AMR path), and another that explores the full AMR structure generated for a sentence. In both cases, we demonstrate that all settings benefit from the fine-grained AMR's semantic structure. We evaluate our model on four RE datasets. Our results show that our model can outperform the GPT-based baselines, and achieve SOTA performance on two of the datasets, and competitive performance on the other two.

Medical Large Language Models (LLMs) such as ClinicalCamel 70B, Llama3-OpenBioLLM 70B have demonstrated impressive performance on a wide variety of medical NLP task.However, there still lacks a large language model (LLM) specifically designed for cancer domain. Moreover, these LLMs typically have billions of parameters, making them computationally expensive for healthcare systems.Thus, in this study, we propose CancerLLM, a model with 7 billion parameters and a Mistral-style architecture, pre-trained on 2,676,642 clinical notes and 515,524 pathology reports covering 17 cancer types, followed by fine-tuning on three cancer-relevant tasks, including cancer phenotypes extraction, cancer diagnosis generation, and cancer treatment plan generation. Our evaluation demonstrated that CancerLLM achieves state-of-the-art results compared to other existing LLMs, with an average F1 score improvement of 8.1\%. Additionally, CancerLLM outperforms other models on two proposed robustness testbeds. This illustrates that CancerLLM can be effectively applied to clinical AI systems, enhancing clinical research and healthcare delivery in the field of cancer.

Evaluation benchmarks are the cornerstone of measuring capabilities of large language models (LLMs), as well as driving progress in said capabilities. Originally designed to make claims about capabilities (or lack thereof) in fully pretrained models, evaluation benchmarks are now also extensively used to decide between various training choices. Despite this widespread usage, we rarely quantify the variance in our evaluation benchmarks, which dictates whether differences in performance are meaningful. Here, we define and measure a range of metrics geared towards measuring variance in evaluation benchmarks, including seed variance across initialisations, and monotonicity during training. By studying a large number of models -- both openly available and pretrained from scratch -- we provide empirical estimates for a variety of variance metrics, with considerations and recommendations for practitioners. We also evaluate the utility and tradeoffs of continuous versus discrete performance measures and explore options for better understanding and reducing this variance. We find that simple changes, such as framing choice tasks (like MMLU) as completion tasks, can often reduce variance for smaller scale ($\sim$7B) models, while more involved methods inspired from human testing literature (such as item analysis and item response theory) struggle to meaningfully reduce variance. Overall, our work provides insights into variance in evaluation benchmarks, suggests LM-specific techniques to reduce variance, and more generally encourages practitioners to carefully factor in variance when comparing models.

Reward models trained on human preference data have been proven to be effective for aligning Large Language Models (LLMs) with human intent within the reinforcement learning from human feedback (RLHF) framework. However, the generalization capabilities of current reward models to unseen prompts and responses are limited. This limitation can lead to an unexpected phenomenon known as reward over-optimization, where excessive optimization of rewards results in a decline in actual performance. While previous research has advocated for constraining policy optimization, our study proposes a novel approach to enhance the reward model's generalization ability against distribution shifts by regularizing the hidden states. Specifically, we retain the base model's language model head and incorporate a suite of text-generation losses to preserve the hidden states' text generation capabilities, while concurrently learning a reward head behind the same hidden states. Our experimental results demonstrate that the introduced regularization technique markedly improves the accuracy of learned reward models across a variety of out-of-distribution (OOD) tasks and effectively alleviate the over-optimization issue in RLHF, offering a more reliable and robust preference learning paradigm.

Graphical User Interface (GUI) automation holds significant promise for enhancing human productivity by assisting with computer tasks. Existing task formulations primarily focus on simple tasks that can be specified by a single, language-only instruction, such as "Insert a new slide." In this work, we introduce VideoGUI, a novel multi-modal benchmark designed to evaluate GUI assistants on visual-centric GUI tasks. Sourced from high-quality web instructional videos, our benchmark focuses on tasks involving professional and novel software (e.g., Adobe Photoshop or Stable Diffusion WebUI) and complex activities (e.g., video editing). VideoGUI evaluates GUI assistants through a hierarchical process, allowing for identification of the specific levels at which they may fail: (i) high-level planning: reconstruct procedural subtasks from visual conditions without language descriptions; (ii) middle-level planning: generate sequences of precise action narrations based on visual state (i.e., screenshot) and goals; (iii) atomic action execution: perform specific actions such as accurately clicking designated elements. For each level, we design evaluation metrics across individual dimensions to provide clear signals, such as individual performance in clicking, dragging, typing, and scrolling for atomic action execution. Our evaluation on VideoGUI reveals that even the SoTA large multimodal model GPT4o performs poorly on visual-centric GUI tasks, especially for high-level planning.

Examining limitations is a crucial step in the scholarly research reviewing process, revealing aspects where a study might lack decisiveness or require enhancement. This aids readers in considering broader implications for further research. In this article, we present a novel and challenging task of Suggestive Limitation Generation (SLG) for research papers. We compile a dataset called \textbf{\textit{LimGen}}, encompassing 4068 research papers and their associated limitations from the ACL anthology. We investigate several approaches to harness large language models (LLMs) for producing suggestive limitations, by thoroughly examining the related challenges, practical insights, and potential opportunities. Our LimGen dataset and code can be accessed at \url{https://github.com/arbmf/LimGen}.

Existing Vision-Language Compositionality (VLC) benchmarks like SugarCrepe are formulated as image-to-text retrieval problems, where, given an image, the models need to select between the correct textual description and a synthetic hard negative text. In this work we present the Bidirectional Vision-Language Compositionality (BiVLC) dataset. The novelty of BiVLC is to add a synthetic hard negative image generated from the synthetic text, resulting in two image-to-text retrieval examples (one for each image) and, more importantly, two text-to-image retrieval examples (one for each text). Human annotators filter out ill-formed examples ensuring the validity of the benchmark. The experiments on BiVLC uncover a weakness of current multimodal models, as they perform poorly in the text-to-image direction. In fact, when considering both retrieval directions, the conclusions obtained in previous works change significantly. In addition to the benchmark, we show that a contrastive model trained using synthetic images and texts improves the state of the art in SugarCrepe and in BiVLC for both retrieval directions. The gap to human performance in BiVLC confirms that Vision-Language Compositionality is still a challenging problem. BiVLC and code are available at https://imirandam.github.io/BiVLC_project_page.

In this technical report, we present CarLLaVA, a Vision Language Model (VLM) for autonomous driving, developed for the CARLA Autonomous Driving Challenge 2.0. CarLLaVA uses the vision encoder of the LLaVA VLM and the LLaMA architecture as backbone, achieving state-of-the-art closed-loop driving performance with only camera input and without the need for complex or expensive labels. Additionally, we show preliminary results on predicting language commentary alongside the driving output. CarLLaVA uses a semi-disentangled output representation of both path predictions and waypoints, getting the advantages of the path for better lateral control and the waypoints for better longitudinal control. We propose an efficient training recipe to train on large driving datasets without wasting compute on easy, trivial data. CarLLaVA ranks 1st place in the sensor track of the CARLA Autonomous Driving Challenge 2.0 outperforming the previous state of the art by 458% and the best concurrent submission by 32.6%.