Safety alignment is an important procedure before the official deployment of a Large Language Model (LLM). While safety alignment has been extensively studied for LLM, there is still a large research gap for Large Reasoning Models (LRMs) that equip with improved reasoning capability. We in this paper systematically examine a simplified pipeline for producing safety aligned LRMs. With our evaluation of various LRMs, we deliver two main findings: i) Safety alignment can be done upon the LRM to restore its safety capability. ii) Safety alignment leads to a degradation of the reasoning capability of LRMs. The two findings show that there exists a trade-off between reasoning and safety capability with the sequential LRM production pipeline. The discovered trade-off, which we name Safety Tax, should shed light on future endeavors of safety research on LRMs. As a by-product, we curate a dataset called DirectRefusal, which might serve as an alternative dataset for safety alignment. Our source code is available at https://github.com/git-disl/Safety-Tax.
Large Language Models (LLMs) and Multi-Agent LLMs (MALLMs) introduce non-determinism unlike traditional or machine learning software, requiring new approaches to verifying correctness beyond simple output comparisons or statistical accuracy over test datasets. This paper presents a taxonomy for LLM test case design, informed by both the research literature, our experience, and open-source tools that represent the state of practice. We identify key variation points that impact test correctness and highlight open challenges that the research, industry, and open-source communities must address as LLMs become integral to software systems. Our taxonomy defines four facets of LLM test case design, addressing ambiguity in both inputs and outputs while establishing best practices. It distinguishes variability in goals, the system under test, and inputs, and introduces two key oracle types: atomic and aggregated. Our mapping indicates that current tools insufficiently account for these variability points, highlighting the need for closer collaboration between academia and practitioners to improve the reliability and reproducibility of LLM testing.
We have seen remarkable success in representation learning and language models (LMs) using deep neural networks. Many studies aim to build the underlying connections among different modalities via the alignment and mappings at the token or embedding level, but so far, most methods are very data-hungry, limiting their performance in domains such as music where paired data are less abundant. We argue that the embedding alignment is only at the surface level of multimodal alignment. In this paper, we propose a grand challenge of \textit{language model mapping} (LMM), i.e., how to map the essence implied in the LM of one domain to the LM of another domain under the assumption that LMs of different modalities are tracking the same underlying phenomena. We first introduce a basic setup of LMM, highlighting the goal to unveil a deeper aspect of cross-modal alignment as well as to achieve more sample-efficiency learning. We then discuss why music is an ideal domain in which to conduct LMM research. After that, we connect LMM in music with a more general and challenging scientific problem of \textit{learning to take actions based on both sensory input and abstract symbols}, and in the end, present an advanced version of the challenge problem setup.
Merging natural language interfaces with computational fluid dynamics (CFD) workflows presents transformative opportunities for both industry and research. In this study, we introduce OptMetaOpenFOAM - a novel framework that bridges MetaOpenFOAM with external analysis and optimization tool libraries through a large language model (LLM)-driven chain-of-thought (COT) methodology. By automating complex CFD tasks via natural language inputs, the framework empowers non-expert users to perform sensitivity analyses and parameter optimizations with markedly improved efficiency. The test dataset comprises 11 distinct CFD analysis or optimization tasks, including a baseline simulation task derived from an OpenFOAM tutorial covering fluid dynamics, combustion, and heat transfer. Results confirm that OptMetaOpenFOAM can accurately interpret user requirements expressed in natural language and effectively invoke external tool libraries alongside MetaOpenFOAM to complete the tasks. Furthermore, validation on a non-OpenFOAM tutorial case - namely, a hydrogen combustion chamber - demonstrates that a mere 200-character natural language input can trigger a sequence of simulation, postprocessing, analysis, and optimization tasks spanning over 2,000 lines of code. These findings underscore the transformative potential of LLM-driven COT methodologies in linking external tool for advanced analysis and optimization, positioning OptMetaOpenFOAM as an effective tool that streamlines CFD simulations and enhances their convenience and efficiency for both industrial and research applications. Code is available at https://github.com/Terry-cyx/MetaOpenFOAM.
Recent years have witnessed a surge in the development of protein structural tokenization methods, which chunk protein 3D structures into discrete or continuous representations. Structure tokenization enables the direct application of powerful techniques like language modeling for protein structures, and large multimodal models to integrate structures with protein sequences and functional texts. Despite the progress, the capabilities and limitations of these methods remain poorly understood due to the lack of a unified evaluation framework. We first introduce StructTokenBench, a framework that comprehensively evaluates the quality and efficiency of structure tokenizers, focusing on fine-grained local substructures rather than global structures, as typical in existing benchmarks. Our evaluations reveal that no single model dominates all benchmarking perspectives. Observations of codebook under-utilization led us to develop AminoAseed, a simple yet effective strategy that enhances codebook gradient updates and optimally balances codebook size and dimension for improved tokenizer utilization and quality. Compared to the leading model ESM3, our method achieves an average of 6.31% performance improvement across 24 supervised tasks, with sensitivity and utilization rates increased by 12.83% and 124.03%, respectively. Source code and model weights are available at https://github.com/KatarinaYuan/StructTokenBench
We introduce InspireMusic, a framework integrated super resolution and large language model for high-fidelity long-form music generation. A unified framework generates high-fidelity music, songs, and audio, which incorporates an autoregressive transformer with a super-resolution flow-matching model. This framework enables the controllable generation of high-fidelity long-form music at a higher sampling rate from both text and audio prompts. Our model differs from previous approaches, as we utilize an audio tokenizer with one codebook that contains richer semantic information, thereby reducing training costs and enhancing efficiency. This combination enables us to achieve high-quality audio generation with long-form coherence of up to $8$ minutes. Then, an autoregressive transformer model based on Qwen 2.5 predicts audio tokens. Next, we employ a super-resolution flow-matching model to generate high-sampling rate audio with fine-grained details learned from an acoustic codec model. Comprehensive experiments show that the InspireMusic-1.5B-Long model has a comparable performance to recent top-tier open-source systems, including MusicGen and Stable Audio 2.0, on subjective and objective evaluations. The code and pre-trained models are released at https://github.com/FunAudioLLM/InspireMusic.
Current studies have exposed the risk of Large Language Models (LLMs) generating harmful content by jailbreak attacks. However, they overlook that the direct generation of harmful content from scratch is more difficult than inducing LLM to calibrate benign content into harmful forms. In our study, we introduce a novel attack framework that exploits AdVersArial meTAphoR (AVATAR) to induce the LLM to calibrate malicious metaphors for jailbreaking. Specifically, to answer harmful queries, AVATAR adaptively identifies a set of benign but logically related metaphors as the initial seed. Then, driven by these metaphors, the target LLM is induced to reason and calibrate about the metaphorical content, thus jailbroken by either directly outputting harmful responses or calibrating residuals between metaphorical and professional harmful content. Experimental results demonstrate that AVATAR can effectively and transferable jailbreak LLMs and achieve a state-of-the-art attack success rate across multiple advanced LLMs.
Large language models are increasingly used in social sciences, but their training data can introduce lookahead bias and training leakage. A good chronologically consistent language model requires efficient use of training data to maintain accuracy despite time-restricted data. Here, we overcome this challenge by training a suite of chronologically consistent large language models, ChronoBERT and ChronoGPT, which incorporate only the text data that would have been available at each point in time. Despite this strict temporal constraint, our models achieve strong performance on natural language processing benchmarks, outperforming or matching widely used models (e.g., BERT), and remain competitive with larger open-weight models. Lookahead bias is model and application-specific because even if a chronologically consistent language model has poorer language comprehension, a regression or prediction model applied on top of the language model can compensate. In an asset pricing application predicting next-day stock returns from financial news, we find that ChronoBERT's real-time outputs achieve a Sharpe ratio comparable to state-of-the-art models, indicating that lookahead bias is modest. Our results demonstrate a scalable, practical framework to mitigate training leakage, ensuring more credible backtests and predictions across finance and other social science domains.
The adaptivity and maneuvering capabilities of Autonomous Underwater Vehicles (AUVs) have drawn significant attention in oceanic research, due to the unpredictable disturbances and strong coupling among the AUV's degrees of freedom. In this paper, we developed large language model (LLM)-enhanced reinforcement learning (RL)-based adaptive S-surface controller for AUVs. Specifically, LLMs are introduced for the joint optimization of controller parameters and reward functions in RL training. Using multi-modal and structured explicit task feedback, LLMs enable joint adjustments, balance multiple objectives, and enhance task-oriented performance and adaptability. In the proposed controller, the RL policy focuses on upper-level tasks, outputting task-oriented high-level commands that the S-surface controller then converts into control signals, ensuring cancellation of nonlinear effects and unpredictable external disturbances in extreme sea conditions. Under extreme sea conditions involving complex terrain, waves, and currents, the proposed controller demonstrates superior performance and adaptability in high-level tasks such as underwater target tracking and data collection, outperforming traditional PID and SMC controllers.
Recent advancements in language models (LMs) have demonstrated strong capabilities in semantic understanding and contextual modeling, which have flourished in generative speech enhancement (SE). However, many LM-based SE approaches primarily focus on semantic information, often neglecting the critical role of acoustic information, which leads to acoustic inconsistency after enhancement and limited generalization across diverse SE tasks. In this paper, we introduce LLaSE-G1, a LLaMA-based language model that incentivizes generalization capabilities for speech enhancement. LLaSE-G1 offers the following key contributions: First, to mitigate acoustic inconsistency, LLaSE-G1 employs continuous representations from WavLM as input and predicts speech tokens from X-Codec2, maximizing acoustic preservation. Second, to promote generalization capability, LLaSE-G1 introduces dual-channel inputs and outputs, unifying multiple SE tasks without requiring task-specific IDs. Third, LLaSE-G1 outperforms prior task-specific discriminative and generative SE models, demonstrating scaling effects at test time and emerging capabilities for unseen SE tasks. Additionally, we release our code and models to support further research in this area.
With the rapid advent of generative models, efficiently deploying these models on specialized hardware has become critical. Tensor Processing Units (TPUs) are designed to accelerate AI workloads, but their high power consumption necessitates innovations for improving efficiency. Compute-in-memory (CIM) has emerged as a promising paradigm with superior area and energy efficiency. In this work, we present a TPU architecture that integrates digital CIM to replace conventional digital systolic arrays in matrix multiply units (MXUs). We first establish a CIM-based TPU architecture model and simulator to evaluate the benefits of CIM for diverse generative model inference. Building upon the observed design insights, we further explore various CIM-based TPU architectural design choices. Up to 44.2% and 33.8% performance improvement for large language model and diffusion transformer inference, and 27.3x reduction in MXU energy consumption can be achieved with different design choices, compared to the baseline TPUv4i architecture.
Large language models (LLMs) hold substantial promise for clinical decision support. However, their widespread adoption in medicine, particularly in healthcare, is hindered by their propensity to generate false or misleading outputs, known as hallucinations. In high-stakes domains such as women's health (obstetrics & gynaecology), where errors in clinical reasoning can have profound consequences for maternal and neonatal outcomes, ensuring the reliability of AI-generated responses is critical. Traditional methods for quantifying uncertainty, such as perplexity, fail to capture meaning-level inconsistencies that lead to misinformation. Here, we evaluate semantic entropy (SE), a novel uncertainty metric that assesses meaning-level variation, to detect hallucinations in AI-generated medical content. Using a clinically validated dataset derived from UK RCOG MRCOG examinations, we compared SE with perplexity in identifying uncertain responses. SE demonstrated superior performance, achieving an AUROC of 0.76 (95% CI: 0.75-0.78), compared to 0.62 (0.60-0.65) for perplexity. Clinical expert validation further confirmed its effectiveness, with SE achieving near-perfect uncertainty discrimination (AUROC: 0.97). While semantic clustering was successful in only 30% of cases, SE remains a valuable tool for improving AI safety in women's health. These findings suggest that SE could enable more reliable AI integration into clinical practice, particularly in resource-limited settings where LLMs could augment care. This study highlights the potential of SE as a key safeguard in the responsible deployment of AI-driven tools in women's health, leading to safer and more effective digital health interventions.
Tuberculosis (TB) is the leading cause of death from an infectious disease globally, with the highest burden in low- and middle-income countries. In these regions, limited healthcare access and high patient-to-provider ratios impede effective patient support, communication, and treatment completion. To bridge this gap, we propose integrating a specialized Large Language Model into an efficacious digital adherence technology to augment interactive communication with treatment supporters. This AI-powered approach, operating within a human-in-the-loop framework, aims to enhance patient engagement and improve TB treatment outcomes.
Mathematical formulas are a fundamental and widely used component in various scientific fields, serving as a universal language for expressing complex concepts and relationships. While state-of-the-art transformer models excel in processing and understanding natural language, they encounter challenges with mathematical notation, which involves a complex structure and diverse representations. This study focuses on the development of specialized training datasets to enhance the encoding of mathematical content. We introduce Math Mutator (MAMUT), a framework capable of generating equivalent and falsified versions of a given mathematical formula in LaTeX notation, effectively capturing the mathematical variety in notation of the same concept. Based on MAMUT, we have generated four large mathematical datasets containing diverse notation, which can be used to train language models with enhanced mathematical embeddings.
EXplainable machine learning (XML) has recently emerged to address the mystery mechanisms of machine learning (ML) systems by interpreting their 'black box' results. Despite the development of various explanation methods, determining the most suitable XML method for specific ML contexts remains unclear, highlighting the need for effective evaluation of explanations. The evaluating capabilities of the Transformer-based large language model (LLM) present an opportunity to adopt LLM-as-a-Judge for assessing explanations. In this paper, we propose a workflow that integrates both LLM-based and human judges for evaluating explanations. We examine how LLM-based judges evaluate the quality of various explanation methods and compare their evaluation capabilities to those of human judges within an iris classification scenario, employing both subjective and objective metrics. We conclude that while LLM-based judges effectively assess the quality of explanations using subjective metrics, they are not yet sufficiently developed to replace human judges in this role.
Strategic decision-making involves interactive reasoning where agents adapt their choices in response to others, yet existing evaluations of large language models (LLMs) often emphasize Nash Equilibrium (NE) approximation, overlooking the mechanisms driving their strategic choices. To bridge this gap, we introduce an evaluation framework grounded in behavioral game theory, disentangling reasoning capability from contextual effects. Testing 22 state-of-the-art LLMs, we find that GPT-o3-mini, GPT-o1, and DeepSeek-R1 dominate most games yet also demonstrate that the model scale alone does not determine performance. In terms of prompting enhancement, Chain-of-Thought (CoT) prompting is not universally effective, as it increases strategic reasoning only for models at certain levels while providing limited gains elsewhere. Additionally, we investigate the impact of encoded demographic features on the models, observing that certain assignments impact the decision-making pattern. For instance, GPT-4o shows stronger strategic reasoning with female traits than males, while Gemma assigns higher reasoning levels to heterosexual identities compared to other sexual orientations, indicating inherent biases. These findings underscore the need for ethical standards and contextual alignment to balance improved reasoning with fairness.
We introduce SEKI, a novel large language model (LLM)-based neural architecture search (NAS) method. Inspired by the chain-of-thought (CoT) paradigm in modern LLMs, SEKI operates in two key stages: self-evolution and knowledge distillation. In the self-evolution stage, LLMs initially lack sufficient reference examples, so we implement an iterative refinement mechanism that enhances architectures based on performance feedback. Over time, this process accumulates a repository of high-performance architectures. In the knowledge distillation stage, LLMs analyze common patterns among these architectures to generate new, optimized designs. Combining these two stages, SEKI greatly leverages the capacity of LLMs on NAS and without requiring any domain-specific data. Experimental results show that SEKI achieves state-of-the-art (SOTA) performance across various datasets and search spaces while requiring only 0.05 GPU-days, outperforming existing methods in both efficiency and accuracy. Furthermore, SEKI demonstrates strong generalization capabilities, achieving SOTA-competitive results across multiple tasks.
Chinese spelling correction (CSC) is a crucial task that aims to correct character errors in Chinese text. While conventional CSC focuses on character substitution errors caused by mistyping, two other common types of character errors, missing and redundant characters, have received less attention. These errors are often excluded from CSC datasets during the annotation process or ignored during evaluation, even when they have been annotated. This issue limits the practicality of the CSC task. To address this issue, we introduce the task of General Chinese Character Error Correction (C2EC), which focuses on all three types of character errors. We construct a high-quality C2EC benchmark by combining and manually verifying data from CCTC and Lemon datasets. We extend the training-free prompt-free CSC method to C2EC by using Levenshtein distance for handling length changes and leveraging an additional prompt-based large language model (LLM) to improve performance. Experiments show that our method enables a 14B-parameter LLM to be on par with models nearly 50 times larger on both conventional CSC and C2EC tasks, without any fine-tuning.
Large language models are increasingly used to process documents and facilitate question-answering on them. In our paper, we extract mechanistic circuits for this real-world language modeling task: context-augmented language modeling for extractive question-answering (QA) tasks and understand the potential benefits of circuits towards downstream applications such as data attribution to context information. We extract circuits as a function of internal model components (e.g., attention heads, MLPs) using causal mediation analysis techniques. Leveraging the extracted circuits, we first understand the interplay between the model's usage of parametric memory and retrieved context towards a better mechanistic understanding of context-augmented language models. We then identify a small set of attention heads in our circuit which performs reliable data attribution by default, thereby obtaining attribution for free in just the model's forward pass. Using this insight, we then introduce ATTNATTRIB, a fast data attribution algorithm which obtains state-of-the-art attribution results across various extractive QA benchmarks. Finally, we show the possibility to steer the language model towards answering from the context, instead of the parametric memory by using the attribution from ATTNATTRIB as an additional signal during the forward pass. Beyond mechanistic understanding, our paper provides tangible applications of circuits in the form of reliable data attribution and model steering.
Robust alignment guardrails for large language models are becoming increasingly important with their widespread application. In contrast to previous studies, we demonstrate that inference-time activation interventions can bypass safety alignments and effectively steer model generations towards harmful AI coordination for Llama 2. Our method applies fine-grained interventions at specific model subcomponents, particularly attention heads, using a simple binary choice probing strategy. These interventions then generalise to the open-ended generation setting effectively circumventing safety guardrails. We show that probing single attention heads is more effective than intervening on full layers and intervening on only four attention heads is comparable to supervised fine-tuning. We further show that only a few example completions are needed to compute effective steering directions, which is an advantage over classical fine-tuning. Our findings highlight the shortcomings of current alignment techniques. In addition, our results suggest that, at the attention head level, activations encode fine-grained linearly separable behaviors. Practically, the approach offers a straightforward methodology to steer large language model behaviour, which could be extended to diverse domains beyond safety requiring fine-grained control over the model output. The code and datasets for this study can be found on https://github.com/PaulDrm/targeted_intervention.