Knowledge-Guided Adversarial Training for Infrared Object Detection via Thermal Radiation Modeling

In complex environments, infrared object detection exhibits broad applicability and stability across diverse scenarios. However, infrared object detection is vulnerable to both common corruptions and adversarial examples, leading to potential security risks. To improve the robustness of infrared object detection, current methods mostly adopt a data-driven ideology, which only superficially drives the network to fit the training data without specifically considering the unique characteristics of infrared images, resulting in limited robustness. In this paper, we revisit infrared physical knowledge and find that relative thermal radiation relations between different classes can be regarded as a reliable knowledge source under the complex scenarios of adversarial examples and common corruptions. Thus, we theoretically model thermal radiation relations based on the rank order of gray values for different classes, and further quantify the stability of various inter-class thermal radiation relations. Based on the above theoretical framework, we propose Knowledge-Guided Adversarial Training (KGAT) for infrared object detection, in which infrared physical knowledge is embedded into the adversarial training process, and the predicted results are optimized to be consistent with the actual physical laws. Extensive experiments on three infrared datasets and six mainstream infrared object detection models demonstrate that KGAT effectively enhances both clean accuracy and robustness against adversarial attacks and common corruptions.

Video-Only ToM: Enhancing Theory of Mind in Multimodal Large Language Models

As large language models (LLMs) continue to advance, there is increasing interest in their ability to infer human mental states and demonstrate a human-like Theory of Mind (ToM). Most existing ToM evaluations, however, are centered on text-based inputs, while scenarios relying solely on visual information receive far less attention. This leaves a gap, since real-world human-AI interaction typically requires multimodal understanding. In addition, many current methods regard the model as a black box and rarely probe how its internal attention behaves in multiple-choice question answering (QA). The impact of LLM hallucinations on such tasks is also underexplored from an interpretability perspective. To address these issues, we introduce VisionToM, a vision-oriented intervention framework designed to strengthen task-aware reasoning. The core idea is to compute intervention vectors that align visual representations with the correct semantic targets, thereby steering the model's attention through different layers of visual features. This guidance reduces the model's reliance on spurious linguistic priors, leading to more reliable multimodal language model (MLLM) outputs and better QA performance. Experiments on the EgoToM benchmark-an egocentric, real-world video dataset for ToM with three multiple-choice QA settings-demonstrate that our method substantially improves the ToM abilities of MLLMs. Furthermore, results on an additional open-ended generation task show that VisionToM enables MLLMs to produce free-form explanations that more accurately capture agents' mental states, pushing machine-human collaboration toward greater alignment.

Step-DeepResearch Technical Report

As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.

XYZCylinder: Feedforward Reconstruction for Driving Scenes Based on A Unified Cylinder Lifting Method

Recently, more attention has been paid to feedforward reconstruction paradigms, which mainly learn a fixed view transformation implicitly and reconstruct the scene with a single representation. However, their generalization capability and reconstruction accuracy are still limited while reconstructing driving scenes, which results from two aspects: (1) The fixed view transformation fails when the camera configuration changes, limiting the generalization capability across different driving scenes equipped with different camera configurations. (2) The small overlapping regions between sparse views of the $360^\circ$ panorama and the complexity of driving scenes increase the learning difficulty, reducing the reconstruction accuracy. To handle these difficulties, we propose \textbf{XYZCylinder}, a feedforward model based on a unified cylinder lifting method which involves camera modeling and feature lifting. Specifically, to improve the generalization capability, we design a Unified Cylinder Camera Modeling (UCCM) strategy, which avoids the learning of viewpoint-dependent spatial correspondence and unifies different camera configurations with adjustable parameters. To improve the reconstruction accuracy, we propose a hybrid representation with several dedicated modules based on newly designed Cylinder Plane Feature Group (CPFG) to lift 2D image features to 3D space. Experimental results show that XYZCylinder achieves state-of-the-art performance under different evaluation settings, and can be generalized to other driving scenes in a zero-shot manner. Project page: \href{https://yuyuyu223.github.io/XYZCYlinder-projectpage/}{here}.

Step-Audio 2 Technical Report

This paper presents Step-Audio 2, an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation. By integrating a latent audio encoder and reasoning-centric reinforcement learning (RL), Step-Audio 2 achieves promising performance in automatic speech recognition (ASR) and audio understanding. To facilitate genuine end-to-end speech conversation, Step-Audio 2 incorporates the generation of discrete audio tokens into language modeling, significantly enhancing its responsiveness to paralinguistic information such as speaking styles and emotions. To effectively leverage the rich textual and acoustic knowledge in real-world data, Step-Audio 2 integrates retrieval-augmented generation (RAG) and is able to call external tools such as web search to mitigate hallucination and audio search to switch timbres. Trained on millions of hours of speech and audio data, Step-Audio 2 delivers intelligence and expressiveness across diverse conversational scenarios. Evaluation results demonstrate that Step-Audio 2 achieves state-of-the-art performance on various audio understanding and conversational benchmarks compared to other open-source and commercial solutions. Please visit https://github.com/stepfun-ai/Step-Audio2 for more information.

From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language Models

As large language models evolve, there is growing anticipation that they will emulate human-like Theory of Mind (ToM) to assist with routine tasks. However, existing methods for evaluating machine ToM focus primarily on unimodal models and largely treat these models as black boxes, lacking an interpretative exploration of their internal mechanisms. In response, this study adopts an approach based on internal mechanisms to provide an interpretability-driven assessment of ToM in multimodal large language models (MLLMs). Specifically, we first construct a multimodal ToM test dataset, GridToM, which incorporates diverse belief testing tasks and perceptual information from multiple perspectives. Next, our analysis shows that attention heads in multimodal large models can distinguish cognitive information across perspectives, providing evidence of ToM capabilities. Furthermore, we present a lightweight, training-free approach that significantly enhances the model's exhibited ToM by adjusting in the direction of the attention head.

Step-Audio-AQAA: a Fully End-to-End Expressive Large Audio Language Model

Large Audio-Language Models (LALMs) have significantly advanced intelligent human-computer interaction, yet their reliance on text-based outputs limits their ability to generate natural speech responses directly, hindering seamless audio interactions. To address this, we introduce Step-Audio-AQAA, a fully end-to-end LALM designed for Audio Query-Audio Answer (AQAA) tasks. The model integrates a dual-codebook audio tokenizer for linguistic and semantic feature extraction, a 130-billion-parameter backbone LLM and a neural vocoder for high-fidelity speech synthesis. Our post-training approach employs interleaved token-output of text and audio to enhance semantic coherence and combines Direct Preference Optimization (DPO) with model merge to improve performance. Evaluations on the StepEval-Audio-360 benchmark demonstrate that Step-Audio-AQAA excels especially in speech control, outperforming the state-of-art LALMs in key areas. This work contributes a promising solution for end-to-end LALMs and highlights the critical role of token-based vocoder in enhancing overall performance for AQAA tasks.

DeCoDe: Defer-and-Complement Decision-Making via Decoupled Concept Bottleneck Models

In human-AI collaboration, a central challenge is deciding whether the AI should handle a task, be deferred to a human expert, or be addressed through collaborative effort. Existing Learning to Defer approaches typically make binary choices between AI and humans, neglecting their complementary strengths. They also lack interpretability, a critical property in high-stakes scenarios where users must understand and, if necessary, correct the model's reasoning. To overcome these limitations, we propose Defer-and-Complement Decision-Making via Decoupled Concept Bottleneck Models (DeCoDe), a concept-driven framework for human-AI collaboration. DeCoDe makes strategy decisions based on human-interpretable concept representations, enhancing transparency throughout the decision process. It supports three flexible modes: autonomous AI prediction, deferral to humans, and human-AI collaborative complementarity, selected via a gating network that takes concept-level inputs and is trained using a novel surrogate loss that balances accuracy and human effort. This approach enables instance-specific, interpretable, and adaptive human-AI collaboration. Experiments on real-world datasets demonstrate that DeCoDe significantly outperforms AI-only, human-only, and traditional deferral baselines, while maintaining strong robustness and interpretability even under noisy expert annotations.

QAVA: Query-Agnostic Visual Attack to Large Vision-Language Models

In typical multimodal tasks, such as Visual Question Answering (VQA), adversarial attacks targeting a specific image and question can lead large vision-language models (LVLMs) to provide incorrect answers. However, it is common for a single image to be associated with multiple questions, and LVLMs may still answer other questions correctly even for an adversarial image attacked by a specific question. To address this, we introduce the query-agnostic visual attack (QAVA), which aims to create robust adversarial examples that generate incorrect responses to unspecified and unknown questions. Compared to traditional adversarial attacks focused on specific images and questions, QAVA significantly enhances the effectiveness and efficiency of attacks on images when the question is unknown, achieving performance comparable to attacks on known target questions. Our research broadens the scope of visual adversarial attacks on LVLMs in practical settings, uncovering previously overlooked vulnerabilities, particularly in the context of visual adversarial threats. The code is available at https://github.com/btzyd/qava.

CIBR: Cross-modal Information Bottleneck Regularization for Robust CLIP Generalization

Contrastive Language-Image Pretraining (CLIP) has achieved remarkable success in cross-modal tasks such as zero-shot image classification and text-image retrieval by effectively aligning visual and textual representations. However, the theoretical foundations underlying CLIP's strong generalization remain unclear. In this work, we address this gap by proposing the Cross-modal Information Bottleneck (CIB) framework. CIB offers a principled interpretation of CLIP's contrastive learning objective as an implicit Information Bottleneck optimization. Under this view, the model maximizes shared cross-modal information while discarding modality-specific redundancies, thereby preserving essential semantic alignment across modalities. Building on this insight, we introduce a Cross-modal Information Bottleneck Regularization (CIBR) method that explicitly enforces these IB principles during training. CIBR introduces a penalty term to discourage modality-specific redundancy, thereby enhancing semantic alignment between image and text features. We validate CIBR on extensive vision-language benchmarks, including zero-shot classification across seven diverse image datasets and text-image retrieval on MSCOCO and Flickr30K. The results show consistent performance gains over standard CLIP. These findings provide the first theoretical understanding of CLIP's generalization through the IB lens. They also demonstrate practical improvements, offering guidance for future cross-modal representation learning.