Test-time scaling (TTS) has been shown to improve the performance of large language models (LLMs) by sampling and aggregating diverse reasoning paths. However, existing research has overlooked a critical issue: selection bias of reasoning strategies during scaling. Specifically, when generating reasoning processes, LLMs tend to follow certain strategies (e.g., algebraic solutions for math problems) while neglecting other valid alternatives (e.g., geometric solutions), resulting in insufficient exploration of the solution space. To further understand the impact of this bias, we present a theoretical analysis that reveals when it undermines the effectiveness of test-time scaling. Motivated by this theoretical insight, we introduce TTS-Uniform, a framework designed to mitigate the selection bias of reasoning strategies. It (i) identifies potential strategies, (ii) uniformly allocates the sampling budget across them, and (iii) filters out unstable strategies prior to aggregation. Experimental results show that TTS-Uniform significantly enhances scaling effectiveness across multiple mainstream LLMs and benchmark datasets. Code is available at https://github.com/zongqianwu/Uniform-TTS. Chain-of-thought (CoT) (Wei et al., 2022; Kojima et al., 2022) enhances the reasoning capabilities of large language models (LLMs) by explicitly unfolding intermediate steps (i.e., reasoning paths) before arriving at the final answer. Building on CoT, test-time scaling (TTS) (Zhang et al., 2025; Ji et al., 2025) further improves performance by sampling and aggregating diverse paths. However, existing TTS research (Wang et al., 2022; Snell et al., 2024) overlooks a critical limitation of CoT, which in turn constrains the effectiveness of scaling. Specifically, when tackling a problem, CoT reasoning tends to follow certain strategies while neglecting other valid alternatives.

Optimizing and refining action execution through exploration and interaction is a promising way for robotic manipulation. However, practical approaches to interaction-driven robotic learning are still underexplored, particularly for long-horizon tasks where sequential decision-making, physical constraints, and perceptual uncertainties pose significant challenges. Motivated by embodied cognition theory, we propose RoboSeek, a framework for embodied action execution that leverages interactive experience to accomplish manipulation tasks. RoboSeek optimizes prior knowledge from high-level perception models through closed-loop training in simulation and achieves robust real-world execution via a real2sim2real transfer pipeline. Specifically, we first replicate real-world environments in simulation using 3D reconstruction to provide visually and physically consistent environments, then we train policies in simulation using reinforcement learning and the cross-entropy method leveraging visual priors. The learned policies are subsequently deployed on real robotic platforms for execution. RoboSeek is hardware-agnostic and is evaluated on multiple robotic platforms across eight long-horizon manipulation tasks involving sequential interactions, tool use, and object handling. Our approach achieves an average success rate of 79%, significantly outperforming baselines whose success rates remain below 50%, highlighting its generalization and robustness across tasks and platforms. Experimental results validate the effectiveness of our training framework in complex, dynamic real-world settings and demonstrate the stability of the proposed real2sim2real transfer mechanism, paving the way for more generalizable embodied robotic learning. Project Page: https://russderrick.github.io/Roboseek/

Standard language models employ unique, monolithic embeddings for each token, potentially limiting their ability to capture the multifaceted nature of word meanings. We investigate whether tokens can be more effectively represented through a compositional structure that accumulates diverse semantic facets. To explore this, we propose Aggregate Semantic Grouping (ASG), a novel approach leveraging Product Quantization (PQ). We apply ASG to standard transformer architectures (mBERT, XLM-R, mT5) and evaluate this representational scheme across diverse tasks (NLI, NER, QA), as well as a biomedical domain-specific benchmark (BC5CDR) using BioBERT. Our findings demonstrate that representing tokens compositionally via ASG achieves extreme compression in embedding parameters (0.4--0.5\%) while maintaining $>$95\% task performance relative to the base model, even in generative tasks and extends to both cross lingual transfer and domain-specific settings. These results validate the principle that tokens can be effectively modeled as combinations of shared semantic building blocks. ASG offers a simple yet concrete method for achieving this, showcasing how compositional representations can capture linguistic richness while enabling compact yet semantically rich models.

The increasing availability of open Earth Observation (EO) and agricultural datasets holds great potential for supporting sustainable land management. However, their high technical entry barrier limits accessibility for non-expert users. This study presents an open-source conversational assistant that integrates multimodal retrieval and large language models (LLMs) to enable natural language interaction with heterogeneous agricultural and geospatial data. The proposed architecture combines orthophotos, Sentinel-2 vegetation indices, and user-provided documents through retrieval-augmented generation (RAG), allowing the system to flexibly determine whether to rely on multimodal evidence, textual knowledge, or both in formulating an answer. To assess response quality, we adopt an LLM-as-a-judge methodology using Qwen3-32B in a zero-shot, unsupervised setting, applying direct scoring in a multi-dimensional quantitative evaluation framework. Preliminary results show that the system is capable of generating clear, relevant, and context-aware responses to agricultural queries, while remaining reproducible and scalable across geographic regions. The primary contributions of this work include an architecture for fusing multimodal EO and textual knowledge sources, a demonstration of lowering the barrier to access specialized agricultural information through natural language interaction, and an open and reproducible design.

The rapid adoption of large language models (LLMs) in critical domains has spurred extensive research into their security issues. While input manipulation attacks (e.g., prompt injection) have been well studied, Bit-Flip Attacks (BFAs) -- which exploit hardware vulnerabilities to corrupt model parameters and cause severe performance degradation -- have received far less attention. Existing BFA methods suffer from key limitations: they fail to balance performance degradation and output naturalness, making them prone to discovery. In this paper, we introduce SilentStriker, the first stealthy bit-flip attack against LLMs that effectively degrades task performance while maintaining output naturalness. Our core contribution lies in addressing the challenge of designing effective loss functions for LLMs with variable output length and the vast output space. Unlike prior approaches that rely on output perplexity for attack loss formulation, which inevitably degrade output naturalness, we reformulate the attack objective by leveraging key output tokens as targets for suppression, enabling effective joint optimization of attack effectiveness and stealthiness. Additionally, we employ an iterative, progressive search strategy to maximize attack efficacy. Experiments show that SilentStriker significantly outperforms existing baselines, achieving successful attacks without compromising the naturalness of generated text.

Existing table understanding methods face challenges due to complex table structures and intricate logical reasoning. While supervised finetuning (SFT) dominates existing research, reinforcement learning (RL), such as Group Relative Policy Optimization (GRPO), has shown promise but struggled with low initial policy accuracy and coarse rewards in tabular contexts. In this paper, we introduce Table-R1, a three-stage RL framework that enhances multimodal table understanding through: (1) Warm-up that prompts initial perception and reasoning capabilities, (2) Perception Alignment GRPO (PA-GRPO), which employs continuous Tree-Edit-Distance Similarity (TEDS) rewards for recognizing table structures and contents, and (3) Hint-Completion GRPO (HC-GRPO), which utilizes fine-grained rewards of residual steps based on the hint-guided question. Extensive experiments demonstrate that Table-R1 can boost the model's table reasoning performance obviously on both held-in and held-out datasets, outperforming SFT and GRPO largely. Notably, Qwen2-VL-7B with Table-R1 surpasses larger specific table understanding models (e.g., Table-LLaVA 13B), even achieving comparable performance to the closed-source model GPT-4o on held-in datasets, demonstrating the efficacy of each stage of Table-R1 in overcoming initialization bottlenecks and reward sparsity, thereby advancing robust multimodal table understanding.

To mitigate the hallucination problem in large language models, DoLa exploits early exit logits from the same model as a contrastive prior. However, we found that these early exit logits tend to be flat, low in magnitude, and fail to reflect meaningful contrasts. To address this, we propose PruneCD, a novel contrastive decoding method that constructs the amateur model via layer pruning rather than early exit. This design leads to more informative and well-aligned logits, enabling more effective contrastive decoding. Through qualitative and quantitative analyses, we demonstrate that PruneCD consistently improves factuality with minimal inference overhead, offering a robust and practical approach to mitigating hallucinations in LLMs.

In this paper, we introduce a novel weighted co-training approach that is guided by Large Language Models (LLMs). Namely, in our co-training approach, we use LLM labels on unlabeled data as target labels and co-train two encoder-only based networks that train each other over multiple iterations: first, all samples are forwarded through each network and historical estimates of each network's confidence in the LLM label are recorded; second, a dynamic importance weight is derived for each sample according to each network's belief in the quality of the LLM label for that sample; finally, the two networks exchange importance weights with each other -- each network back-propagates all samples weighted with the importance weights coming from its peer network and updates its own parameters. By strategically utilizing LLM-generated guidance, our approach significantly outperforms conventional SSL methods, particularly in settings with abundant unlabeled data. Empirical results show that it achieves state-of-the-art performance on 4 out of 5 benchmark datasets and ranks first among 14 compared methods according to the Friedman test. Our results highlight a new direction in semi-supervised learning -- where LLMs serve as knowledge amplifiers, enabling backbone co-training models to achieve state-of-the-art performance efficiently.

Large Language Models are increasingly used in applications requiring objective assessment, which could be compromised by political bias. Many studies found preferences for left-leaning positions in LLMs, but downstream effects on tasks like fact-checking remain underexplored. In this study, we systematically investigate political bias through exchanging words with euphemisms or dysphemisms in German claims. We construct minimal pairs of factually equivalent claims that differ in political connotation, to assess the consistency of LLMs in classifying them as true or false. We evaluate six LLMs and find that, more than political leaning, the presence of judgmental words significantly influences truthfulness assessment. While a few models show tendencies of political bias, this is not mitigated by explicitly calling for objectivism in prompts. Warning: This paper contains content that may be offensive or upsetting.

Embodied AI development significantly lags behind large foundation models due to three critical challenges: (1) lack of systematic understanding of core capabilities needed for Embodied AI, making research lack clear objectives; (2) absence of unified and standardized evaluation systems, rendering cross-benchmark evaluation infeasible; and (3) underdeveloped automated and scalable acquisition methods for embodied data, creating critical bottlenecks for model scaling. To address these obstacles, we present Embodied Arena, a comprehensive, unified, and evolving evaluation platform for Embodied AI. Our platform establishes a systematic embodied capability taxonomy spanning three levels (perception, reasoning, task execution), seven core capabilities, and 25 fine-grained dimensions, enabling unified evaluation with systematic research objectives. We introduce a standardized evaluation system built upon unified infrastructure supporting flexible integration of 22 diverse benchmarks across three domains (2D/3D Embodied Q&A, Navigation, Task Planning) and 30+ advanced models from 20+ worldwide institutes. Additionally, we develop a novel LLM-driven automated generation pipeline ensuring scalable embodied evaluation data with continuous evolution for diversity and comprehensiveness. Embodied Arena publishes three real-time leaderboards (Embodied Q&A, Navigation, Task Planning) with dual perspectives (benchmark view and capability view), providing comprehensive overviews of advanced model capabilities. Especially, we present nine findings summarized from the evaluation results on the leaderboards of Embodied Arena. This helps to establish clear research veins and pinpoint critical research problems, thereby driving forward progress in the field of Embodied AI.