The advent of test-time scaling in large language models (LLMs), exemplified by OpenAI's o1 series, has advanced reasoning capabilities by scaling computational resource allocation during inference. While successors like QwQ, Deepseek-R1 (R1) and LIMO replicate these advancements, whether these models truly possess test-time scaling capabilities remains underexplored. This study found that longer CoTs of these o1-like models do not consistently enhance accuracy; in fact, correct solutions are often shorter than incorrect ones for the same questions. Further investigation shows this phenomenon is closely related to models' self-revision capabilities - longer CoTs contain more self-revisions, which often lead to performance degradation. We then compare sequential and parallel scaling strategies on QwQ, R1 and LIMO, finding that parallel scaling achieves better coverage and scalability. Based on these insights, we propose "Shortest Majority Vote", a method that combines parallel scaling strategies with CoT length characteristics, significantly improving models' test-time scalability compared to conventional majority voting approaches.

Humans exhibit remarkable compositional reasoning by integrating knowledge from various sources. For example, if someone learns ( B = f(A) ) from one source and ( C = g(B) ) from another, they can deduce ( C=g(B)=g(f(A)) ) even without encountering ( ABC ) together, showcasing the generalization ability of human intelligence. In this paper, we introduce a synthetic learning task, "FTCT" (Fragmented at Training, Chained at Testing), to validate the potential of Transformers in replicating this skill and interpret its inner mechanism. In the training phase, data consist of separated knowledge fragments from an overall causal graph. During testing, Transformers must infer complete causal graph traces by integrating these fragments. Our findings demonstrate that few-shot Chain-of-Thought prompting enables Transformers to perform compositional reasoning on FTCT by revealing correct combinations of fragments, even if such combinations were absent in the training data. Furthermore, the emergence of compositional reasoning ability is strongly correlated with the model complexity and training-testing data similarity. We propose, both theoretically and empirically, that Transformers learn an underlying generalizable program from training, enabling effective compositional reasoning during testing.

Retrieval-augmented language models often struggle with knowledge-intensive tasks due to inefficient retrieval, unstructured knowledge integration, and single-pass architectures. We present Retrieval-And-Structuring (RAS), a novel framework that dynamically constructs and reasons over query-specific knowledge graphs through iterative retrieval and structuring. RAS introduces four key technical innovations: (1) a themescoped retrieval mechanism that efficiently narrows the search space while maintaining retrieval quality, (2) an action planning module that determines knowledge needs and generates focused sub-queries, (3) a dynamic knowledge structuring approach that converts retrieved text into an evolving knowledge graph, and (4) a graph-augmented answering component that leverages the accumulated structured information. Our framework achieves state-of-the-art performance, surpassing leading baselines by 6.4% with open-source language models and 7.0% with proprietary models on seven knowledge-intensive generation datasets across all evaluation metrics. Detailed ablation studies verify the contribution of each technical component to the overall system performance.

Recently, Long Chain-of-Thoughts (CoTs) have gained widespread attention for improving the reasoning capabilities of Large Language Models (LLMs). This necessitates that existing LLMs, which lack the ability to generate Long CoTs, to acquire such capability through post-training methods. Without additional training, LLMs typically enhance their mathematical reasoning abilities through inference scaling methods such as MCTS. However, they are hindered by the large action space and inefficient search strategies, making it challenging to generate Long CoTs effectively. To tackle this issue, we propose constraining the action space and guiding the emergence of Long CoTs through a refined search strategy. In our proposed Constrained Monte Carlo Tree Search (C-MCTS) framework, we limit the actions selected from a constrained action space, which is divided into five disjoint subsets: \emph{understanding}, \emph{planning}, \emph{reflection}, \emph{coding}, and \emph{summary}. Each subset is further constrained to a small number of predefined prompts, rather than allowing LLMs to generate actions arbitrarily. Additionally, we refine the search strategy by incorporating prior knowledge about the action sets, such as a human-like partial order of the action subsets and the pretrained process reward models. These strategies work together to significantly reduce the vast search space of Long CoTs. Extensive evaluations on mathematical reasoning benchmarks show that, under zero-shot settings, our method enables the 7B model to achieve reasoning capabilities that surpass those of the 72B model.

Our objective in this paper is to develop a machinery that makes a given organizational strategic plan resilient to the actions of competitor agents (adverse environmental actions). We assume that we are given a goal tree representing strategic goals (can also be seen business requirements for a software systems) with the assumption that competitor agents are behaving in a maximally adversarial fashion(opposing actions against our sub goals or goals in general). We use game tree search methods (such as minimax) to select an optimal execution strategy(at a given point in time), such that it can maximize our chances of achieving our (high level) strategic goals. Our machinery helps us determine which path to follow(strategy selection) to achieve the best end outcome. This is done by comparing alternative execution strategies available to us via an evaluation function. Our evaluation function is based on the idea that we want to make our execution plans defensible(future-proof) by selecting execution strategies that make us least vulnerable to adversarial actions by the competitor agents.

Longer thought, better performance: large language models with deep reasoning capabilities, particularly o1-like models, have demonstrated remarkable performance by generating extensive thought processes during inference. This trade-off reveals a potential vulnerability: adversaries could compromise model performance by forcing immediate responses without thought processes. To this end, in this paper, we introduce a novel attack scenario targeting the long thought processes of o1-like models and propose BoT (Break CoT), which can selectively break intrinsic reasoning mechanisms through backdoor attacks. BoT constructs poisoned datasets with designed triggers and injects backdoor by either supervised fine-tuning or direct preference optimization. When triggered, the model directly generates answers without thought processes, while maintaining normal reasoning capabilities for clean inputs. Extensive experiments on open-source o1-like models, including recent DeepSeek-R1, demonstrate that BoT nearly achieves high attack success rates while maintaining clean accuracy, highlighting the critical safety risk in current models. Furthermore, the relationship between task difficulty and helpfulness reveals a potential application for good, enabling users to customize model behavior based on task complexity. Code is available at \href{https://github.com/zihao-ai/BoT}{https://github.com/zihao-ai/BoT}.

Entity tracking is a fundamental challenge in natural language understanding, requiring models to maintain coherent representations of entities. Previous work has benchmarked entity tracking performance in purely text-based tasks. We introduce MET-Bench, a multimodal entity tracking benchmark designed to evaluate the ability of vision-language models to track entity states across modalities. Using two structured domains, Chess and the Shell Game, we assess how effectively current models integrate textual and image-based state updates. Our findings reveal a significant performance gap between text-based and image-based tracking and that this performance gap stems from deficits in visual reasoning rather than perception. We further show that explicit text-based reasoning strategies improve performance, yet substantial limitations remain, especially in long-horizon multimodal scenarios. Our results highlight the need for improved multimodal representations and reasoning techniques to bridge the gap between textual and visual entity tracking.

Recent breakthroughs in autonomous driving have been propelled by advances in robust world modeling, fundamentally transforming how vehicles interpret dynamic scenes and execute safe decision-making. In particular, world models have emerged as a linchpin technology, offering high-fidelity representations of the driving environment that integrate multi-sensor data, semantic cues, and temporal dynamics. This paper systematically reviews recent advances in world models for autonomous driving, proposing a three-tiered taxonomy: 1) Generation of Future Physical World, covering image-, BEV-, OG-, and PC-based generation methods that enhance scene evolution modeling through diffusion models and 4D occupancy forecasting; 2) Behavior Planning for Intelligent Agents, combining rule-driven and learning-based paradigms with cost map optimization and reinforcement learning for trajectory generation in complex traffic conditions; 3) Interaction Between Prediction and Planning, achieving multi-agent collaborative decision-making through latent space diffusion and memory-augmented architectures. The study further analyzes training paradigms including self-supervised learning, multimodal pretraining, and generative data augmentation, while evaluating world models' performance in scene understanding and motion prediction tasks. Future research must address key challenges in self-supervised representation learning, long-tail scenario generation, and multimodal fusion to advance the practical deployment of world models in complex urban environments. Overall, our comprehensive analysis provides a theoretical framework and technical roadmap for harnessing the transformative potential of world models in advancing safe and reliable autonomous driving solutions.

OpenAI o1 has shown that applying reinforcement learning to integrate reasoning steps directly during inference can significantly improve a model's reasoning capabilities. This result is exciting as the field transitions from the conventional autoregressive method of generating answers to a more deliberate approach that models the slow-thinking process through step-by-step reasoning training. Reinforcement learning plays a key role in both the model's training and decoding processes. In this article, we present a comprehensive formulation of reasoning problems and investigate the use of both model-based and model-free approaches to better support this slow-thinking framework.

We consider the problem of test-time domain generalization, where a model is trained on several source domains and adjusted on target domains never seen during training. Different from the common methods that fine-tune the model or adjust the classifier parameters online, we propose to generate multiple layer parameters on the fly during inference by a lightweight meta-learned transformer, which we call \textit{GeneralizeFormer}. The layer-wise parameters are generated per target batch without fine-tuning or online adjustment. By doing so, our method is more effective in dynamic scenarios with multiple target distributions and also avoids forgetting valuable source distribution characteristics. Moreover, by considering layer-wise gradients, the proposed method adapts itself to various distribution shifts. To reduce the computational and time cost, we fix the convolutional parameters while only generating parameters of the Batch Normalization layers and the linear classifier. Experiments on six widely used domain generalization datasets demonstrate the benefits and abilities of the proposed method to efficiently handle various distribution shifts, generalize in dynamic scenarios, and avoid forgetting.