Deterministically controlling the target generation language of large multilingual language models (LLMs) remains a fundamental challenge, particularly in zero-shot settings where neither explicit language prompts nor fine-tuning are available. In this work, we investigate whether sparse autoencoder (SAE) features, previously shown to correlate with interpretable model behaviors, can be leveraged to steer the generated language of LLMs during inference. Leveraging pretrained SAEs on the residual streams of Gemma-2B and Gemma-9B, we identify features whose activations differ most significantly between English and four target languages: Chinese, Japanese, Spanish, and French. By modifying just a single SAE feature at one transformer layer, we achieve controlled language shifts with up to 90\% success, as measured by FastText language classification, while preserving semantic fidelity according to LaBSE (Language-Agnostic BERT Sentence Embedding) similarity. Our analysis reveals that language steering is most effective in mid-to-late transformer layers and is amplified by specific attention heads disproportionately associated with language-sensitive SAE features. These results demonstrate the promise of sparse feature steering as a lightweight and interpretable mechanism for controllable multilingual generation.

Objective Renal cancer is a common malignancy and a major cause of cancer-related deaths. Computed tomography (CT) is central to early detection, staging, and treatment planning. However, the growing CT workload increases radiologists' burden and risks incomplete documentation. Automatically generating accurate reports remains challenging because it requires integrating visual interpretation with clinical reasoning. Advances in artificial intelligence (AI), especially large language and vision-language models, offer potential to reduce workload and enhance diagnostic quality. Methods We propose a clinically informed, two-stage framework for automatic renal CT report generation. In Stage 1, a multi-task learning model detects structured clinical features from each 2D image. In Stage 2, a vision-language model generates free-text reports conditioned on the image and the detected features. To evaluate clinical fidelity, generated clinical features are extracted from the reports and compared with expert-annotated ground truth. Results Experiments on an expert-labeled dataset show that incorporating detected features improves both report quality and clinical accuracy. The model achieved an average AUC of 0.75 for key imaging features and a METEOR score of 0.33, demonstrating higher clinical consistency and fewer template-driven errors. Conclusion Linking structured feature detection with conditioned report generation provides a clinically grounded approach to integrate structured prediction and narrative drafting for renal CT reporting. This method enhances interpretability and clinical faithfulness, underscoring the value of domain-relevant evaluation metrics for medical AI development.

Characterizing a large language model's (LLM's) knowledge of a given question is challenging. As a result, prior work has primarily examined LLM behavior under knowledge conflicts, where the model's internal parametric memory contradicts information in the external context. However, this does not fully reflect how well the model knows the answer to the question. In this paper, we first introduce a taxonomy of five knowledge statuses based on the consistency and correctness of LLM knowledge modes. We then propose KScope, a hierarchical framework of statistical tests that progressively refines hypotheses about knowledge modes and characterizes LLM knowledge into one of these five statuses. We apply KScope to nine LLMs across four datasets and systematically establish: (1) Supporting context narrows knowledge gaps across models. (2) Context features related to difficulty, relevance, and familiarity drive successful knowledge updates. (3) LLMs exhibit similar feature preferences when partially correct or conflicted, but diverge sharply when consistently wrong. (4) Context summarization constrained by our feature analysis, together with enhanced credibility, further improves update effectiveness and generalizes across LLMs.

Large language models (LLMs) can be prompted with specific styles (e.g., formatting responses as lists), including in malicious queries. Prior jailbreak research mainly augments these queries with additional string transformations to maximize attack success rate (ASR). However, the impact of style patterns in the original queries that are semantically irrelevant to the malicious intent remains unclear. In this work, we seek to understand whether style patterns compromise LLM safety, how superficial style alignment increases model vulnerability, and how best to mitigate these risks during alignment. We first define ASR inflation as the increase in ASR due to style patterns in existing jailbreak benchmark queries. By evaluating 32 LLMs across seven benchmarks, we find that nearly all models exhibit ASR inflation. Notably, the inflation correlates with an LLM's relative attention to style patterns, which also overlap more with its instruction-tuning data when inflation occurs. We then investigate superficial style alignment, and find that fine-tuning with specific styles makes LLMs more vulnerable to jailbreaks of those same styles. Finally, we propose SafeStyle, a defense strategy that incorporates a small amount of safety training data augmented to match the distribution of style patterns in the fine-tuning data. Across three LLMs, six fine-tuning style settings, and two real-world instruction-tuning datasets, SafeStyle consistently outperforms baselines in maintaining LLM safety.

Fine-tuning Large Language Models (LLMs) is essential for adapting pre-trained models to downstream tasks. Yet traditional first-order optimizers such as Stochastic Gradient Descent (SGD) and Adam incur prohibitive memory and computational costs that scale poorly with model size. In this paper, we investigate zero-order (ZO) optimization methods as a memory- and compute-efficient alternative, particularly in the context of parameter-efficient fine-tuning techniques like LoRA. We propose $\texttt{JAGUAR SignSGD}$, a ZO momentum-based algorithm that extends ZO SignSGD, requiring the same number of parameters as the standard ZO SGD and only $\mathcal{O}(1)$ function evaluations per iteration. To the best of our knowledge, this is the first study to establish rigorous convergence guarantees for SignSGD in the stochastic ZO case. We further propose $\texttt{JAGUAR Muon}$, a novel ZO extension of the Muon optimizer that leverages the matrix structure of model parameters, and we provide its convergence rate under arbitrary stochastic noise. Through extensive experiments on challenging LLM fine-tuning benchmarks, we demonstrate that the proposed algorithms meet or exceed the convergence quality of standard first-order methods, achieving significant memory reduction. Our theoretical and empirical results establish new ZO optimization methods as a practical and theoretically grounded approach for resource-constrained LLM adaptation. Our code is available at https://github.com/brain-mmo-lab/ZO_LLM

Model distillation has become essential for creating smaller, deployable language models that retain larger system capabilities. However, widespread deployment raises concerns about resilience to adversarial manipulation. This paper investigates vulnerability of distilled models to adversarial injection of biased content during training. We demonstrate that adversaries can inject subtle biases into teacher models through minimal data poisoning, which propagates to student models and becomes significantly amplified. We propose two propagation modes: Untargeted Propagation, where bias affects multiple tasks, and Targeted Propagation, focusing on specific tasks while maintaining normal behavior elsewhere. With only 25 poisoned samples (0.25% poisoning rate), student models generate biased responses 76.9% of the time in targeted scenarios - higher than 69.4% in teacher models. For untargeted propagation, adversarial bias appears 6x-29x more frequently in student models on unseen tasks. We validate findings across six bias types (targeted advertisements, phishing links, narrative manipulations, insecure coding practices), various distillation methods, and different modalities spanning text and code generation. Our evaluation reveals shortcomings in current defenses - perplexity filtering, bias detection systems, and LLM-based autorater frameworks - against these attacks. Results expose significant security vulnerabilities in distilled models, highlighting need for specialized safeguards. We propose practical design principles for building effective adversarial bias mitigation strategies.

Cross-lingual transfer is central to modern NLP, enabling models to perform tasks in languages different from those they were trained on. A common assumption is that training on more languages improves zero-shot transfer. We test this on sense-aware tasks-polysemy and lexical semantic change-and find that multilinguality is not necessary for effective transfer. Our large-scale analysis across 28 languages reveals that other factors, such as differences in pretraining and fine-tuning data and evaluation artifacts, better explain the perceived benefits of multilinguality. We also release fine-tuned models and provide empirical baselines to support future research. While focused on two sense-aware tasks, our findings offer broader insights into cross-lingual transfer, especially for low-resource languages.

Recent studies show that the reasoning capabilities of Large Language Models (LLMs) can be improved by applying Reinforcement Learning (RL) to question-answering (QA) tasks in areas such as math and coding. With a long context length, LLMs may learn to perform search, as indicated by the self-correction behavior observed in DeepSeek R1. However, this search behavior is often imprecise and lacks confidence, resulting in long, redundant responses and highlighting deficiencies in intuition and verification. Inspired by the Dual Process Theory in psychology, we introduce a simple modification to the QA task that includes four stages: Fast Thinking, where the LLM must answer within a strict token budget; Verification, where the model evaluates its initial response; Slow Thinking, where it refines the initial response with more deliberation; and Summarization, where it distills the refinement from the previous stage into precise steps. Our proposed task improves average accuracy from 25.6% to 27.3% for Qwen2.5-1.5B, and from 45.9% to 51.0% for DeepSeek-R1-Qwen-1.5B. Notably, for Qwen2.5-1.5B, the Fast Thinking mode alone achieves 25.2% accuracy using fewer than 1000 tokens, demonstrating substantial inference efficiency gains. These findings suggest that intuition and deliberative reasoning are distinct, complementary systems benefiting from targeted training. Additionally, we have open-sourced both the trained models and the source code.

Given the central role of charts in scientific, business, and communication contexts, enhancing the chart understanding capabilities of vision-language models (VLMs) has become increasingly critical. A key limitation of existing VLMs lies in their inaccurate visual grounding of infographic elements, including charts and human-recognizable objects (HROs) such as icons and images. However, chart understanding often requires identifying relevant elements and reasoning over them. To address this limitation, we introduce InfoDet, a dataset designed to support the development of accurate object detection models for charts and HROs in infographics. It contains 11,264 real and 90,000 synthetic infographics, with over 14 million bounding box annotations. These annotations are created by combining the model-in-the-loop and programmatic methods. We demonstrate the usefulness of InfoDet through three applications: 1) constructing a Thinking-with-Boxes scheme to boost the chart understanding performance of VLMs, 2) comparing existing object detection models, and 3) applying the developed detection model to document layout and UI element detection.

As Text-to-Image (TTI) diffusion models become increasingly influential in content creation, growing attention is being directed toward their societal and cultural implications. While prior research has primarily examined demographic and cultural biases, the ability of these models to accurately represent historical contexts remains largely underexplored. To address this gap, we introduce a benchmark for evaluating how TTI models depict historical contexts. The benchmark combines HistVis, a dataset of 30,000 synthetic images generated by three state-of-the-art diffusion models from carefully designed prompts covering universal human activities across multiple historical periods, with a reproducible evaluation protocol. We evaluate generated imagery across three key aspects: (1) Implicit Stylistic Associations: examining default visual styles associated with specific eras; (2) Historical Consistency: identifying anachronisms such as modern artifacts in pre-modern contexts; and (3) Demographic Representation: comparing generated racial and gender distributions against historically plausible baselines. Our findings reveal systematic inaccuracies in historically themed generated imagery, as TTI models frequently stereotype past eras by incorporating unstated stylistic cues, introduce anachronisms, and fail to reflect plausible demographic patterns. By providing a reproducible benchmark for historical representation in generated imagery, this work provides an initial step toward building more historically accurate TTI models.