Evaluating the performance of Grammatical Error Correction (GEC) models has become increasingly challenging, as large language model (LLM)-based GEC systems often produce corrections that diverge from provided gold references. This discrepancy undermines the reliability of traditional reference-based evaluation metrics. In this study, we propose a novel evaluation framework for GEC models, DSGram, integrating Semantic Coherence, Edit Level, and Fluency, and utilizing a dynamic weighting mechanism. Our framework employs the Analytic Hierarchy Process (AHP) in conjunction with large language models to ascertain the relative importance of various evaluation criteria. Additionally, we develop a dataset incorporating human annotations and LLM-simulated sentences to validate our algorithms and fine-tune more cost-effective models. Experimental results indicate that our proposed approach enhances the effectiveness of GEC model evaluations.
Editing complex visual content based on ambiguous instructions remains a challenging problem in vision-language modeling. While existing models can contextualize content, they often struggle to grasp the underlying intent within a reference image or scene, leading to misaligned edits. We introduce the Editing Vision-Language Model (EVLM), a system designed to interpret such instructions in conjunction with reference visuals, producing precise and context-aware editing prompts. Leveraging Chain-of-Thought (CoT) reasoning and KL-Divergence Target Optimization (KTO) alignment technique, EVLM captures subjective editing preferences without requiring binary labels. Fine-tuned on a dataset of 30,000 CoT examples, with rationale paths rated by human evaluators, EVLM demonstrates substantial improvements in alignment with human intentions. Experiments across image, video, 3D, and 4D editing tasks show that EVLM generates coherent, high-quality instructions, supporting a scalable framework for complex vision-language applications.
Recently, with the emergence of recent Multimodal Large Language Model (MLLM) technology, it has become possible to exploit its video understanding capability on different classification tasks. In practice, we face the difficulty of huge requirements for GPU resource if we need to deploy MLLMs online. In this paper, we propose COEF-VQ, a novel cascaded MLLM framework for better video quality understanding on TikTok. To this end, we first propose a MLLM fusing all visual, textual and audio signals, and then develop a cascade framework with a lightweight model as pre-filtering stage and MLLM as fine-consideration stage, significantly reducing the need for GPU resource, while retaining the performance demonstrated solely by MLLM. To demonstrate the effectiveness of COEF-VQ, we deployed this new framework onto the video management platform (VMP) at TikTok, and performed a series of detailed experiments on two in-house tasks related to video quality understanding. We show that COEF-VQ leads to substantial performance gains with limit resource consumption in these two tasks.
This paper introduces a novel approach to creating adaptive language agents by integrating active inference with large language models (LLMs). While LLMs demonstrate remarkable capabilities, their reliance on static prompts limits adaptation to new information and changing environments. We address this by implementing an active inference framework that acts as a cognitive layer above an LLM-based agent, dynamically adjusting prompts and search strategies through principled information-seeking behavior. Our framework models the environment using three state factors (prompt, search, and information states) with seven observation modalities capturing quality metrics. By framing the agent's learning through the free energy principle, we enable systematic exploration of prompt combinations and search strategies. Experimental results demonstrate the effectiveness of this approach, with the agent developing accurate models of environment dynamics evidenced by emergent structure in observation matrices. Action selection patterns reveal sophisticated exploration-exploitation behavior, transitioning from initial information-gathering to targeted prompt testing. The integration of thermodynamic principles with language model capabilities provides a principled framework for creating robust, adaptable agents, extending active inference beyond traditional low-dimensional control problems to high-dimensional, language-driven environments.
Motivation: Codon optimization of Open Reading Frame (ORF) sequences is essential for enhancing mRNA stability and expression in applications like mRNA vaccines, where codon choice can significantly impact protein yield which directly impacts immune strength. In this work, we investigate the use of a pre-trained protein language model (PPLM) for getting a rich representation of amino acids which could be utilized for codon optimization. This leaves us with a simpler fine-tuning task over PPLM in optimizing ORF sequences. Results: The ORFs generated by our proposed models outperformed their natural counterparts encoding the same proteins on computational metrics for stability and expression. They also demonstrated enhanced performance against the benchmark ORFs used in mRNA vaccines for the SARS-CoV-2 viral spike protein and the varicella-zoster virus (VZV). These results highlight the potential of adapting PPLM for designing ORFs tailored to encode target antigens in mRNA vaccines.
We address the problem of interactive text-to-image (T2I) generation, designing a reinforcement learning (RL) agent which iteratively improves a set of generated images for a user through a sequence of prompt expansions. Using human raters, we create a novel dataset of sequential preferences, which we leverage, together with large-scale open-source (non-sequential) datasets. We construct user-preference and user-choice models using an EM strategy and identify varying user preference types. We then leverage a large multimodal language model (LMM) and a value-based RL approach to suggest an adaptive and diverse slate of prompt expansions to the user. Our Preference Adaptive and Sequential Text-to-image Agent (PASTA) extends T2I models with adaptive multi-turn capabilities, fostering collaborative co-creation and addressing uncertainty or underspecification in a user's intent. We evaluate PASTA using human raters, showing significant improvement compared to baseline methods. We also open-source our sequential rater dataset and simulated user-rater interactions to support future research in user-centric multi-turn T2I systems.
Composed Image Retrieval (CIR) aims to retrieve target images from candidate set using a hybrid-modality query consisting of a reference image and a relative caption that describes the user intent. Recent studies attempt to utilize Vision-Language Pre-training Models (VLPMs) with various fusion strategies for addressing the task.However, these methods typically fail to simultaneously meet two key requirements of CIR: comprehensively extracting visual information and faithfully following the user intent. In this work, we propose CIR-LVLM, a novel framework that leverages the large vision-language model (LVLM) as the powerful user intent-aware encoder to better meet these requirements. Our motivation is to explore the advanced reasoning and instruction-following capabilities of LVLM for accurately understanding and responding the user intent. Furthermore, we design a novel hybrid intent instruction module to provide explicit intent guidance at two levels: (1) The task prompt clarifies the task requirement and assists the model in discerning user intent at the task level. (2) The instance-specific soft prompt, which is adaptively selected from the learnable prompt pool, enables the model to better comprehend the user intent at the instance level compared to a universal prompt for all instances. CIR-LVLM achieves state-of-the-art performance across three prominent benchmarks with acceptable inference efficiency. We believe this study provides fundamental insights into CIR-related fields.
Word-by-word language model surprisal is often used to model the incremental processing of human readers, which raises questions about how various choices in language modeling influence its predictive power. One factor that has been overlooked in cognitive modeling is the granularity of subword tokens, which explicitly encodes information about word length and frequency, and ultimately influences the quality of vector representations that are learned. This paper presents experiments that manipulate the token granularity and evaluate its impact on the ability of surprisal to account for processing difficulty of naturalistic text and garden-path constructions. Experiments with naturalistic reading times reveal a substantial influence of token granularity on surprisal, with tokens defined by a vocabulary size of 8,000 resulting in surprisal that is most predictive. In contrast, on garden-path constructions, language models trained on coarser-grained tokens generally assigned higher surprisal to critical regions, suggesting their increased sensitivity to syntax. Taken together, these results suggest a large role of token granularity on the quality of language model surprisal for cognitive modeling.
Mathematical reasoning, a core aspect of human cognition, is vital across many domains, from educational problem-solving to scientific advancements. As artificial general intelligence (AGI) progresses, integrating large language models (LLMs) with mathematical reasoning tasks is becoming increasingly significant. This survey provides the first comprehensive analysis of mathematical reasoning in the era of multimodal large language models (MLLMs). We review over 200 studies published since 2021, and examine the state-of-the-art developments in Math-LLMs, with a focus on multimodal settings. We categorize the field into three dimensions: benchmarks, methodologies, and challenges. In particular, we explore multimodal mathematical reasoning pipeline, as well as the role of (M)LLMs and the associated methodologies. Finally, we identify five major challenges hindering the realization of AGI in this domain, offering insights into the future direction for enhancing multimodal reasoning capabilities. This survey serves as a critical resource for the research community in advancing the capabilities of LLMs to tackle complex multimodal reasoning tasks.
Existing work on large language model (LLM) personalization assigned different responding roles to LLM, but overlooked the diversity of questioners. In this work, we propose a new form of questioner-aware LLM personalization, generating different responses even for the same query from different questioners. We design a dual-tower model architecture with a cross-questioner general encoder and a questioner-specific encoder. We further apply contrastive learning with multi-view augmentation, pulling close the dialogue representations of the same questioner, while pulling apart those of different questioners. To mitigate the impact of question diversity on questioner-contrastive learning, we cluster the dialogues based on question similarity and restrict the scope of contrastive learning within each cluster. We also build a multi-questioner dataset from English and Chinese scripts and WeChat records, called MQDialog, containing 173 questioners and 12 responders. Extensive evaluation with different metrics shows a significant improvement in the quality of personalized response generation.
Legal contracts in the custody and fund services domain govern critical aspects such as key provider responsibilities, fee schedules, and indemnification rights. However, it is challenging for an off-the-shelf Large Language Model (LLM) to ingest these contracts due to the lengthy unstructured streams of text, limited LLM context windows, and complex legal jargon. To address these challenges, we introduce LAW (Legal Agentic Workflows for Custody and Fund Services Contracts). LAW features a modular design that responds to user queries by orchestrating a suite of domain-specific tools and text agents. Our experiments demonstrate that LAW, by integrating multiple specialized agents and tools, significantly outperforms the baseline. LAW excels particularly in complex tasks such as calculating a contract's termination date, surpassing the baseline by 92.9% points. Furthermore, LAW offers a cost-effective alternative to traditional fine-tuned legal LLMs by leveraging reusable, domain-specific tools.
Human-Centric NLP often claims to prioritise human needs and values, yet many implementations reveal an underlying AI-centric focus. Through an analysis of case studies in language modelling, behavioural testing, and multi-modal alignment, this study identifies a significant gap between the ideas of human-centricity and actual practices. Key issues include misalignment with human-centred design principles, the reduction of human factors to mere benchmarks, and insufficient consideration of real-world impacts. The discussion explores whether Human-Centric NLP embodies true human-centred design, emphasising the need for interdisciplinary collaboration and ethical considerations. The paper advocates for a redefinition of Human-Centric NLP, urging a broader focus on real-world utility and societal implications to ensure that language technologies genuinely serve and empower users.
A seminal paper published by Ledley and Lusted in 1959 introduced complex clinical diagnostic reasoning cases as the gold standard for the evaluation of expert medical computing systems, a standard that has held ever since. Here, we report the results of a physician evaluation of a large language model (LLM) on challenging clinical cases against a baseline of hundreds of physicians. We conduct five experiments to measure clinical reasoning across differential diagnosis generation, display of diagnostic reasoning, triage differential diagnosis, probabilistic reasoning, and management reasoning, all adjudicated by physician experts with validated psychometrics. We then report a real-world study comparing human expert and AI second opinions in randomly-selected patients in the emergency room of a major tertiary academic medical center in Boston, MA. We compared LLMs and board-certified physicians at three predefined diagnostic touchpoints: triage in the emergency room, initial evaluation by a physician, and admission to the hospital or intensive care unit. In all experiments--both vignettes and emergency room second opinions--the LLM displayed superhuman diagnostic and reasoning abilities, as well as continued improvement from prior generations of AI clinical decision support. Our study suggests that LLMs have achieved superhuman performance on general medical diagnostic and management reasoning, fulfilling the vision put forth by Ledley and Lusted, and motivating the urgent need for prospective trials.
The rapid advancement of autonomous web navigation has significantly benefited from grounding pretrained Large Language Models (LLMs) as agents. However, current research has yet to fully leverage the redundancy of HTML elements for contrastive training. This paper introduces a novel approach to LLM-based web navigation tasks, called Web Element Preference Optimization (WEPO). WEPO utilizes unsupervised preference learning by sampling distance-based non-salient web elements as negative samples, optimizing maximum likelihood objective within Direct Preference Optimization (DPO). We evaluate WEPO on the Mind2Web benchmark and empirically demonstrate that WEPO aligns user high-level intent with output actions more effectively. The results show that our method achieved the state-of-the-art, with an improvement of 13.8% over WebAgent and 5.3% over the visual language model CogAgent baseline. Our findings underscore the potential of preference optimization to enhance web navigation and other web page based tasks, suggesting a promising direction for future research.
Training automatic summary fact verifiers often faces the challenge of a lack of human-labeled data. In this paper, we explore alternative way of leveraging Large Language Model (LLM) generated feedback to address the inherent limitation of using human-labeled data. We introduce FineSumFact, a large-scale dataset containing fine-grained factual feedback on summaries. We employ 10 distinct LLMs for diverse summary generation and Llama-3-70B-Instruct for feedback. We utilize this dataset to fine-tune the lightweight open-source model Llama-3-8B-Instruct, optimizing resource efficiency while maintaining high performance. Our experimental results reveal that the model trained on extensive LLM-generated datasets surpasses that trained on smaller human-annotated datasets when evaluated using human-generated test sets. Fine-tuning fact verification models with LLM feedback can be more effective and cost-efficient than using human feedback. The dataset is available at https://github.com/DISL-Lab/FineSumFact.
Retrieval-augmented generation (RAG) has emerged as a popular approach to steering the output of a large language model (LLM) by incorporating retrieved contexts as inputs. However, existing work observed the generator bias, such that improving the retrieval results may negatively affect the outcome. In this work, we show such bias can be mitigated, from inference scaling, aggregating inference calls from the permuted order of retrieved contexts. The proposed Mixture-of-Intervention (MOI) explicitly models the debiased utility of each passage with multiple forward passes to construct a new ranking. We also show that MOI can leverage the retriever's prior knowledge to reduce the computational cost by minimizing the number of permutations considered and lowering the cost per LLM call. We showcase the effectiveness of MOI on diverse RAG tasks, improving ROUGE-L on MS MARCO and EM on HotpotQA benchmarks by ~7 points.
WHAT-IF -- Writing a Hero's Alternate Timeline through Interactive Fiction -- is a system that uses zero-shot meta-prompting to create branching narratives from a prewritten story. Played as an interactive fiction (IF) game, WHAT-IF lets the player choose between decisions that the large language model (LLM) GPT-4 generates as possible branches in the story. Starting with an existing linear plot as input, a branch is created at each key decision taken by the main character. By meta-prompting the LLM to consider the major plot points from the story, the system produces coherent and well-structured alternate storylines. WHAT-IF stores the branching plot tree in a graph which helps it to both keep track of the story for prompting and maintain the structure for the final IF system. A video demo of our system can be found here: https://youtu.be/8vBqjqtupcc.
Collecting high-quality preference datasets for reinforcement learning from human feedback (RLHF) is resource-intensive and challenging. As a result, researchers often train reward models on extensive offline datasets which aggregate diverse generation sources and scoring/alignment policies. We hypothesize that this aggregation has an averaging effect on reward model scores, which limits signal and impairs the alignment process. Inspired by the field of inverse RL, we define the 'inverse alignment problem' in language model training, where our objective is to optimize the critic's reward for a fixed actor and a fixed offline preference dataset. We hypothesize that solving the inverse alignment problem will improve reward model quality by providing clearer feedback on the policy's current behavior. To that end, we investigate whether repeatedly fine-tuning a reward model on subsets of the offline preference dataset aligned with a periodically frozen policy during RLHF improves upon vanilla RLHF. Our empirical results demonstrate that this approach facilitates superior alignment and faster convergence compared to using an unaligned or out-of-distribution reward model relative to the LLM policy.
Language model approaches have recently been integrated into binary analysis tasks, such as function similarity detection and function signature recovery. These models typically employ a two-stage training process: pre-training via Masked Language Modeling (MLM) on machine code and fine-tuning for specific tasks. While MLM helps to understand binary code structures, it ignores essential code characteristics, including control and data flow, which negatively affect model generalization. Recent work leverages domain-specific features (e.g., control flow graphs and dynamic execution traces) in transformer-based approaches to improve binary code semantic understanding. However, this approach involves complex feature engineering, a cumbersome and time-consuming process that can introduce predictive uncertainty when dealing with stripped or obfuscated code, leading to a performance drop. In this paper, we introduce ProTST, a novel transformer-based methodology for binary code embedding. ProTST employs a hierarchical training process based on a unique tree-like structure, where knowledge progressively flows from fundamental tasks at the root to more specialized tasks at the leaves. This progressive teacher-student paradigm allows the model to build upon previously learned knowledge, resulting in high-quality embeddings that can be effectively leveraged for diverse downstream binary analysis tasks. The effectiveness of ProTST is evaluated in seven binary analysis tasks, and the results show that ProTST yields an average validation score (F1, MRR, and Recall@1) improvement of 14.8% compared to traditional two-stage training and an average validation score of 10.7% compared to multimodal two-stage frameworks.
Foresight 2 (FS2) is a large language model fine-tuned on hospital data for modelling patient timelines (GitHub 'removed for anon'). It can understand patients' clinical notes and predict SNOMED codes for a wide range of biomedical use cases, including diagnosis suggestions, risk forecasting, and procedure and medication recommendations. FS2 is trained on the free text portion of the MIMIC-III dataset, firstly through extracting biomedical concepts and then creating contextualised patient timelines, upon which the model is then fine-tuned. The results show significant improvement over the previous state-of-the-art for the next new biomedical concept prediction (P/R - 0.73/0.66 vs 0.52/0.32) and a similar improvement specifically for the next new disorder prediction (P/R - 0.69/0.62 vs 0.46/0.25). Finally, on the task of risk forecast, we compare our model to GPT-4-turbo (and a range of open-source biomedical LLMs) and show that FS2 performs significantly better on such tasks (P@5 - 0.90 vs 0.65). This highlights the need to incorporate hospital data into LLMs and shows that small models outperform much larger ones when fine-tuned on high-quality, specialised data.