Temporal claims, often riddled with inaccuracies, are a significant challenge in the digital misinformation landscape. Fact-checking systems that can accurately verify such claims are crucial for combating misinformation. Current systems struggle with the complexities of evaluating the accuracy of these claims, especially when they include multiple, overlapping, or recurring events. We introduce a novel timeline-based fact verification framework that identify events from both claim and evidence and organize them into their respective chronological timelines. The framework systematically examines the relationships between the events in both claim and evidence to predict the veracity of each claim event and their chronological accuracy. This allows us to accurately determine the overall veracity of the claim. We also introduce a new dataset of complex temporal claims involving timeline-based reasoning for the training and evaluation of our proposed framework. Experimental results demonstrate the effectiveness of our approach in handling the intricacies of temporal claim verification.
This study introduces BrainTransformers, an innovative Large Language Model (LLM) implemented using Spiking Neural Networks (SNN). Our key contributions include: (1) designing SNN-compatible Transformer components such as SNNMatmul, SNNSoftmax, and SNNSiLU; (2) implementing an SNN approximation of the SiLU activation function; and (3) developing a Synapsis module to simulate synaptic plasticity. Our 3-billion parameter model, BrainTransformers-3B-Chat, demonstrates competitive performance across various benchmarks, including MMLU (63.2), BBH (54.1), ARC-C (54.3), and GSM8K (76.3), while potentially offering improved energy efficiency and biological plausibility. The model employs a three-stage training approach, including SNN-specific neuronal synaptic plasticity training. This research opens new avenues for brain-like AI systems in natural language processing and neuromorphic computing. Future work will focus on hardware optimization, developing specialized SNN fine-tuning tools, and exploring practical applications in energy-efficient computing environments.
Pretrained language models have shown strong effectiveness in code-related tasks, such as code retrieval, code generation, code summarization, and code completion tasks. In this paper, we propose COde assistaNt viA retrieval-augmeNted language model (CONAN), which aims to build a code assistant by mimicking the knowledge-seeking behaviors of humans during coding. Specifically, it consists of a code structure aware retriever (CONAN-R) and a dual-view code representation-based retrieval-augmented generation model (CONAN-G). CONAN-R pretrains CodeT5 using Code-Documentation Alignment and Masked Entity Prediction tasks to make language models code structure-aware and learn effective representations for code snippets and documentation. Then CONAN-G designs a dual-view code representation mechanism for implementing a retrieval-augmented code generation model. CONAN-G regards the code documentation descriptions as prompts, which help language models better understand the code semantics. Our experiments show that CONAN achieves convincing performance on different code generation tasks and significantly outperforms previous retrieval augmented code generation models. Our further analyses show that CONAN learns tailored representations for both code snippets and documentation by aligning code-documentation data pairs and capturing structural semantics by masking and predicting entities in the code data. Additionally, the retrieved code snippets and documentation provide necessary information from both program language and natural language to assist the code generation process. CONAN can also be used as an assistant for Large Language Models (LLMs), providing LLMs with external knowledge in shorter code document lengths to improve their effectiveness on various code tasks. It shows the ability of CONAN to extract necessary information and help filter out the noise from retrieved code documents.
Benchmark contamination refers to the presence of test datasets in Large Language Model (LLM) pre-training or post-training data. Contamination can lead to inflated scores on benchmarks, compromising evaluation results and making it difficult to determine the capabilities of models. In this work, we study the contamination of popular multilingual benchmarks in LLMs that support multiple languages. We use the Black Box test to determine whether $7$ frequently used multilingual benchmarks are contaminated in $7$ popular open and closed LLMs and find that almost all models show signs of being contaminated with almost all the benchmarks we test. Our findings can help the community determine the best set of benchmarks to use for multilingual evaluation.
Reward models are critical in techniques like Reinforcement Learning from Human Feedback (RLHF) and Inference Scaling Laws, where they guide language model alignment and select optimal responses. Despite their importance, existing reward model benchmarks often evaluate models by asking them to distinguish between responses generated by models of varying power. However, this approach fails to assess reward models on subtle but critical content changes and variations in style, resulting in a low correlation with policy model performance. To this end, we introduce RM-Bench, a novel benchmark designed to evaluate reward models based on their sensitivity to subtle content differences and resistance to style biases. Extensive experiments demonstrate that RM-Bench strongly correlates with policy model performance, making it a reliable reference for selecting reward models to align language models effectively. We evaluate nearly 40 reward models on RM-Bench. Our results reveal that even state-of-the-art models achieve an average performance of only 46.6%, which falls short of random-level accuracy (50%) when faced with style bias interference. These findings highlight the significant room for improvement in current reward models. Related code and data are available at https://github.com/THU-KEG/RM-Bench.
Representation learning of Text-Attributed Graphs (TAGs) has garnered significant attention due to its applications in various domains, including recommendation systems and social networks. Despite advancements in TAG learning methodologies, challenges remain in explainability due to the black-box nature of existing TAG representation learning models. This paper presents TAGExplainer, the first method designed to generate natural language explanations for TAG learning. TAGExplainer employs a generative language model that maps input-output pairs to explanations reflecting the model's decision-making process. To address the lack of annotated ground truth explanations in real-world scenarios, we propose first generating pseudo-labels that capture the model's decisions from saliency-based explanations, then the pseudo-label generator is iteratively trained based on three training objectives focusing on faithfulness and brevity via Expert Iteration, to improve the quality of generated pseudo-labels. The high-quality pseudo-labels are finally utilized to train an end-to-end explanation generator model. Extensive experiments are conducted to demonstrate the effectiveness of TAGExplainer in producing faithful and concise natural language explanations.
Intermediate task transfer learning can greatly improve model performance. If, for example, one has little training data for emotion detection, first fine-tuning a language model on a sentiment classification dataset may improve performance strongly. But which task to choose for transfer learning? Prior methods producing useful task rankings are infeasible for large source pools, as they require forward passes through all source language models. We overcome this by introducing Embedding Space Maps (ESMs), light-weight neural networks that approximate the effect of fine-tuning a language model. We conduct the largest study on NLP task transferability and task selection with 12k source-target pairs. We find that applying ESMs on a prior method reduces execution time and disk space usage by factors of 10 and 278, respectively, while retaining high selection performance (avg. regret@5 score of 2.95).
A plethora of jailbreaking attacks have been proposed to obtain harmful responses from safety-tuned LLMs. In their original settings, these methods all largely succeed in coercing the target output, but their attacks vary substantially in fluency and computational effort. In this work, we propose a unified threat model for the principled comparison of these methods. Our threat model combines constraints in perplexity, measuring how far a jailbreak deviates from natural text, and computational budget, in total FLOPs. For the former, we build an N-gram model on 1T tokens, which, in contrast to model-based perplexity, allows for an LLM-agnostic and inherently interpretable evaluation. We adapt popular attacks to this new, realistic threat model, with which we, for the first time, benchmark these attacks on equal footing. After a rigorous comparison, we not only find attack success rates against safety-tuned modern models to be lower than previously presented but also find that attacks based on discrete optimization significantly outperform recent LLM-based attacks. Being inherently interpretable, our threat model allows for a comprehensive analysis and comparison of jailbreak attacks. We find that effective attacks exploit and abuse infrequent N-grams, either selecting N-grams absent from real-world text or rare ones, e.g. specific to code datasets.
We introduce the Seismic Language Model (SeisLM), a foundational model designed to analyze seismic waveforms -- signals generated by Earth's vibrations such as the ones originating from earthquakes. SeisLM is pretrained on a large collection of open-source seismic datasets using a self-supervised contrastive loss, akin to BERT in language modeling. This approach allows the model to learn general seismic waveform patterns from unlabeled data without being tied to specific downstream tasks. When fine-tuned, SeisLM excels in seismological tasks like event detection, phase-picking, onset time regression, and foreshock-aftershock classification. The code has been made publicly available on https://github.com/liutianlin0121/seisLM.
KV cache compression methods have mainly relied on scalar quantization techniques to reduce the memory requirements during decoding. In this work, we apply residual vector quantization, which has been widely used for high fidelity audio compression, to compress KV cache in large language models (LLM). We adapt the standard recipe with minimal changes to compress the output of any key or value projection matrix in a pretrained LLM: we scale the vector by its standard deviation, divide channels into groups and then quantize each group with the same residual vector quantizer. We learn the codebook using exponential moving average and there are no other learnable parameters including the input and output projections normally used in a vector quantization set up. We find that a residual depth of 8 recovers most of the performance of the unquantized model. We also find that grouping non-contiguous channels together works better than grouping contiguous channels for compressing key matrix and the method further benefits from a light weight finetuning of LLM together with the quantization. Overall, the proposed technique is competitive with existing quantization methods while being much simpler and results in 5.5x compression compared to half precision.
Protein structures are important for understanding their functions and interactions. Currently, many protein structure prediction methods are enriching the structure database. Discriminating the origin of structures is crucial for distinguishing between experimentally resolved and computationally predicted structures, evaluating the reliability of prediction methods, and guiding downstream biological studies. Building on works in structure prediction, We developed a structure-sensitive supervised deep learning model, Crystal vs Predicted Evaluator for Protein Structure (CPE-Pro), to represent and discriminate the origin of protein structures. CPE-Pro learns the structural information of proteins and captures inter-structural differences to achieve accurate traceability on four data classes, and is expected to be extended to more. Simultaneously, we utilized Foldseek to encode protein structures into "structure-sequences" and trained a protein Structural Sequence Language Model, SSLM. Preliminary experiments demonstrated that, compared to large-scale protein language models pre-trained on vast amounts of amino acid sequences, the "structure-sequence" enables the language model to learn more informative protein features, enhancing and optimizing structural representations. We have provided the code, model weights, and all related materials on https://github.com/GouWenrui/CPE-Pro-main.git.
Large Language Models (LLMs) are thought to struggle with arithmetic learning due to the inherent differences between language modeling and numerical computation, but concrete evidence has been lacking. This work responds to this claim through a two-side experiment. We first investigate whether LLMs leverage partial products during arithmetic learning. We find that although LLMs can identify some partial products after learning, they fail to leverage them for arithmetic tasks, conversely. We then explore how LLMs approach arithmetic symbolically by breaking tasks into subgroups, hypothesizing that difficulties arise from subgroup complexity and selection. Our results show that when subgroup complexity is fixed, LLMs treat a collection of different arithmetic operations similarly. By analyzing position-level accuracy across different training sizes, we further observe that it follows a U-shaped pattern: LLMs quickly learn the easiest patterns at the first and last positions, while progressively learning the more difficult patterns in the middle positions. This suggests that LLMs select subgroup following an easy-to-hard paradigm during learning. Our work confirms that LLMs are pure symbolic learners in arithmetic tasks and underscores the importance of understanding them deeply through subgroup-level quantification.
The Transformer architecture has become widely adopted due to its demonstrated success, attributed to the attention mechanism at its core. Despite these successes, the attention mechanism of Transformers is associated with two well-known issues: rank-collapse and gradient vanishing. In this paper, we present a theoretical analysis that it is inherently difficult to address both issues simultaneously in the conventional attention mechanism. To handle these issues, we introduce a novel class of attention mechanism, referred to as generalized probabilistic attention mechanism (GPAM), and its dual-attention implementation within the Transformer architecture. Unlike conventional attention mechanisms, GPAM allows for negative attention scores while preserving a fixed total sum. We provide theoretical evidence that the proposed dual-attention GPAM (daGPAM) effectively mitigates both the rank-collapse and gradient vanishing issues which are difficult to resolve simultaneously with the conventional attention mechanisms. Furthermore, we empirically validate this theoretical evidence, demonstrating the superiority of daGPAM compared to other alternative attention mechanisms that were proposed to address the same issues. Additionally, we demonstrate the practical benefits of GPAM in natural language processing tasks, such as language modeling and neural machine translation.
Video generation models have made significant progress in simulating future states, showcasing their potential as world simulators in embodied scenarios. However, existing models often lack robust understanding, limiting their ability to perform multi-step predictions or handle Out-of-Distribution (OOD) scenarios. To address this challenge, we propose the Reflection of Generation (RoG), a set of intermediate reasoning strategies designed to enhance video prediction. It leverages the complementary strengths of pre-trained vision-language and video generation models, enabling them to function as a world model in embodied scenarios. To support RoG, we introduce Embodied Video Anticipation Benchmark(EVA-Bench), a comprehensive benchmark that evaluates embodied world models across diverse tasks and scenarios, utilizing both in-domain and OOD datasets. Building on this foundation, we devise a world model, Embodied Video Anticipator (EVA), that follows a multistage training paradigm to generate high-fidelity video frames and apply an autoregressive strategy to enable adaptive generalization for longer video sequences. Extensive experiments demonstrate the efficacy of EVA in various downstream tasks like video generation and robotics, thereby paving the way for large-scale pre-trained models in real-world video prediction applications. The video demos are available at \hyperlink{https://sites.google.com/view/icml-eva}{https://sites.google.com/view/icml-eva}.
We propose a zero-shot text-driven 3D shape deformation system that deforms an input 3D mesh of a manufactured object to fit an input text description. To do this, our system optimizes the parameters of a deformation model to maximize an objective function based on the widely used pre-trained vision language model CLIP. We find that CLIP-based objective functions exhibit many spurious local optima; to circumvent them, we parameterize deformations using a novel deformation model called BoxDefGraph which our system automatically computes from an input mesh, the BoxDefGraph is designed to capture the object aligned rectangular/circular geometry features of most manufactured objects. We then use the CMA-ES global optimization algorithm to maximize our objective, which we find to work better than popular gradient-based optimizers. We demonstrate that our approach produces appealing results and outperforms several baselines.
We develop a method that integrates the tree of thoughts and multi-agent framework to enhance the capability of pre-trained language models in solving complex, unfamiliar games. The method decomposes game-solving into four incremental tasks -- game summarization, area selection, action extraction, and action validation -- each assigned to a specific language-model agent. By constructing a tree of thoughts, the method simulates reasoning paths and allows agents to collaboratively distill game representations and tactics, mitigating the limitations of language models in reasoning and long-term memorization. Additionally, an automated fine-tuning process further optimizes the agents' performance by ranking query-response pairs based on game outcomes, e.g., winning or losing. We apply the method to a non-cooperative game and demonstrate a 65 percent winning rate against benchmark algorithms, with an additional 10 percent improvement after fine-tuning. In contrast to existing deep learning algorithms for game solving that require millions of training samples, the proposed method consumes approximately 1000 training samples, highlighting its efficiency and scalability.
Although pre-trained language models can generate executable plans (e.g., programmatic policies) for solving robot tasks, the generated plans may violate task-relevant logical specifications due to the models' black-box nature. A significant gap remains between the language models' outputs and verifiable executions of plans. We develop a method to generate executable plans and formally verify them against task-relevant safety specifications. Given a high-level task description in natural language, the proposed method queries a language model to generate plans in the form of executable robot programs. It then converts the generated plan into an automaton-based representation, allowing formal verification of the automaton against the specifications. We prove that given a set of verified plans, the composition of these plans also satisfies the safety specifications. This proof ensures the safety of complex, multi-component plans, obviating the computation complexity of verifying the composed plan. We then propose an automated fine-tuning process that refines the language model to generate specification-compliant plans without the need for human labeling. The empirical results show a 30 percent improvement in the probability of generating plans that meet task specifications after fine-tuning.
Developing language model-based dialogue agents requires effective data to train models that can follow specific task logic. However, most existing data augmentation methods focus on increasing diversity in language, topics, or dialogue acts at the utterance level, largely neglecting a critical aspect of task logic diversity at the dialogue level. This paper proposes a novel data augmentation method designed to enhance the diversity of synthetic dialogues by focusing on task execution logic. Our method uses LLMs to generate decision tree-structured task plans, which enables the derivation of diverse dialogue trajectories for a given task. Each trajectory, referred to as a "dialog flow", guides the generation of a multi-turn dialogue that follows a unique trajectory. We apply this method to generate a task-oriented dialogue dataset comprising 3,886 dialogue flows across 15 different domains. We validate the effectiveness of this dataset using the next action prediction task, where models fine-tuned on our dataset outperform strong baselines, including GPT-4. Upon acceptance of this paper, we plan to release the code and data publicly.
Large Language Models (LLMs) have shown impressive capabilities in many scenarios, but their performance depends, in part, on the choice of prompt. Past research has focused on optimizing prompts specific to a task. However, much less attention has been given to optimizing the general instructions included in a prompt, known as a system prompt. To address this gap, we propose SPRIG, an edit-based genetic algorithm that iteratively constructs prompts from prespecified components to maximize the model's performance in general scenarios. We evaluate the performance of system prompts on a collection of 47 different types of tasks to ensure generalizability. Our study finds that a single optimized system prompt performs on par with task prompts optimized for each individual task. Moreover, combining system and task-level optimizations leads to further improvement, which showcases their complementary nature. Experiments also reveal that the optimized system prompts generalize effectively across model families, parameter sizes, and languages. This study provides insights into the role of system-level instructions in maximizing LLM potential.
We present MMDS, a system capable of recognizing medical images and patient facial details, and providing professional medical diagnoses. The system consists of two core components:The first component is the analysis of medical images and videos. We trained a specialized multimodal medical model capable of interpreting medical images and accurately analyzing patients' facial emotions and facial paralysis conditions. The model achieved an accuracy of 72.59% on the FER2013 facial emotion recognition dataset, with a 91.1% accuracy in recognizing the "happy" emotion. In facial paralysis recognition, the model reached an accuracy of 92%, which is 30% higher than that of GPT-4o. Based on this model, we developed a parser for analyzing facial movement videos of patients with facial paralysis, achieving precise grading of the paralysis severity. In tests on 30 videos of facial paralysis patients, the system demonstrated a grading accuracy of 83.3%.The second component is the generation of professional medical responses. We employed a large language model, integrated with a medical knowledge base, to generate professional diagnoses based on the analysis of medical images or videos. The core innovation lies in our development of a department-specific knowledge base routing management mechanism, in which the large language model categorizes data by medical departments and, during the retrieval process, determines the appropriate knowledge base to query. This significantly improves retrieval accuracy in the RAG (retrieval-augmented generation) process.