In autonomous driving, dynamic environment and corner cases pose significant challenges to the robustness of ego vehicle's decision-making. To address these challenges, commencing with the representation of state-action mapping in the end-to-end autonomous driving paradigm, we introduce a novel pipeline, VDT-Auto. Leveraging the advancement of the state understanding of Visual Language Model (VLM), incorporating with diffusion Transformer-based action generation, our VDT-Auto parses the environment geometrically and contextually for the conditioning of the diffusion process. Geometrically, we use a bird's-eye view (BEV) encoder to extract feature grids from the surrounding images. Contextually, the structured output of our fine-tuned VLM is processed into textual embeddings and noisy paths. During our diffusion process, the added noise for the forward process is sampled from the noisy path output of the fine-tuned VLM, while the extracted BEV feature grids and embedded texts condition the reverse process of our diffusion Transformers. Our VDT-Auto achieved 0.52m on average L2 errors and 21% on average collision rate in the nuScenes open-loop planning evaluation. Moreover, the real-world demonstration exhibited prominent generalizability of our VDT-Auto. The code and dataset will be released after acceptance.

Large Language Models (LLMs) have recently shown remarkable advancement in various NLP tasks. As such, a popular trend has emerged lately where NLP researchers extract word/sentence/document embeddings from these large decoder-only models and use them for various inference tasks with promising results. However, it is still unclear whether the performance improvement of LLM-induced embeddings is merely because of scale or whether underlying embeddings they produce significantly differ from classical encoding models like Word2Vec, GloVe, Sentence-BERT (SBERT) or Universal Sentence Encoder (USE). This is the central question we investigate in the paper by systematically comparing classical decontextualized and contextualized word embeddings with the same for LLM-induced embeddings. Our results show that LLMs cluster semantically related words more tightly and perform better on analogy tasks in decontextualized settings. However, in contextualized settings, classical models like SimCSE often outperform LLMs in sentence-level similarity assessment tasks, highlighting their continued relevance for fine-grained semantics.

In recent years, with advances in large language models (LLMs), end-to-end spoken dialogue models (SDMs) have made significant strides. Compared to text-based LLMs, the evaluation of SDMs needs to take speech-related aspects into account, such as paralinguistic information and speech quality. However, there is still a lack of comprehensive evaluations for SDMs in speech-to-speech (S2S) scenarios. To address this gap, we propose URO-Bench, an extensive benchmark for SDMs. Notably, URO-Bench is the first S2S benchmark that covers evaluations about multilingualism, multi-round dialogues, and paralinguistics. Our benchmark is divided into two difficulty levels: basic track and pro track, consisting of 16 and 20 datasets respectively, evaluating the model's abilities in Understanding, Reasoning, and Oral conversation. Evaluations on our proposed benchmark reveal that current open-source SDMs perform rather well in daily QA tasks, but lag behind their backbone LLMs in terms of instruction-following ability and also suffer from catastrophic forgetting. Their performance in advanced evaluations of paralinguistic information and audio understanding remains subpar, highlighting the need for further research in this direction. We hope that URO-Bench can effectively facilitate the development of spoken dialogue models by providing a multifaceted evaluation of existing models and helping to track progress in this area.

Reward models (RMs) are crucial for aligning large language models (LLMs) with human preferences. However, most RM research is centered on English and relies heavily on synthetic resources, which leads to limited and less reliable datasets and benchmarks for Chinese. To address this gap, we introduce CheemsBench, a fully human-annotated RM evaluation benchmark within Chinese contexts, and CheemsPreference, a large-scale and diverse preference dataset annotated through human-machine collaboration to support Chinese RM training. We systematically evaluate open-source discriminative and generative RMs on CheemsBench and observe significant limitations in their ability to capture human preferences in Chinese scenarios. Additionally, based on CheemsPreference, we construct an RM that achieves state-of-the-art performance on CheemsBench, demonstrating the necessity of human supervision in RM training. Our findings reveal that scaled AI-generated data struggles to fully capture human preferences, emphasizing the importance of high-quality human supervision in RM development.

Speculative decoding is a powerful technique that accelerates Large Language Model (LLM) inference by leveraging a lightweight speculative draft model. However, existing designs suffers in performance due to misalignment between training and inference. Recent methods have tried to solve this issue by adopting a multi-step training strategy, but the complex inputs of different training steps make it harder for the draft model to converge. To address this, we propose CORAL, a novel framework that improves both accuracy and efficiency in speculative drafting. CORAL introduces Cross-Step Representation Alignment, a method that enhances consistency across multiple training steps, significantly improving speculative drafting performance. Additionally, we identify the LM head as a major bottleneck in the inference speed of the draft model. We introduce a weight-grouping mechanism that selectively activates a subset of LM head parameters during inference, substantially reducing the latency of the draft model. We evaluate CORAL on three LLM families and three benchmark datasets, achieving speedup ratios of 2.50x-4.07x, outperforming state-of-the-art methods such as EAGLE-2 and HASS. Our results demonstrate that CORAL effectively mitigates training-inference misalignment and delivers significant speedup for modern LLMs with large vocabularies.

Large Language Models (LLMs) achieve state-of-the-art performance across various NLP tasks but face deployment challenges due to high computational costs and memory constraints. Knowledge distillation (KD) is a promising solution, transferring knowledge from large teacher models to smaller student models. However, existing KD methods often assume shared vocabularies and tokenizers, limiting their flexibility. While approaches like Universal Logit Distillation (ULD) and Dual-Space Knowledge Distillation (DSKD) address vocabulary mismatches, they overlook the critical \textbf{reasoning-aware distillation} aspect. To bridge this gap, we propose CoT2Align a universal KD framework that integrates Chain-of-Thought (CoT) augmentation and introduces Cross-CoT Alignment to enhance reasoning transfer. Additionally, we extend Optimal Transport beyond token-wise alignment to a sequence-level and layer-wise alignment approach that adapts to varying sequence lengths while preserving contextual integrity. Comprehensive experiments demonstrate that CoT2Align outperforms existing KD methods across different vocabulary settings, improving reasoning capabilities and robustness in domain-specific tasks.

Recently, Devin has made a significant buzz in the Artificial Intelligence (AI) community as the world's first fully autonomous AI software engineer, capable of independently developing software code [1] [2]. Devin uses the concept of agentic workflow in Generative AI (GenAI), which empowers AI agents to engage in a more dynamic, iterative, and self-reflective process. With Saarthi, verification engineers can focus on more complex problems, and verification teams can strive for more ambitious goals. The domain-agnostic implementation of Saarthi makes it scalable for use across various domains such as RTL design, UVM-based verification, and others. Hardware design verification, especially formal verification, entails a methodical and disciplined approach to the planning, development, execution, and sign-off of functionally correct hardware designs. Formal verification uses mathematical methods to prove the correctness of hardware designs against their specifications, ensuring that all possible states and inputs are considered, which complements traditional simulation-based verification techniques that might only cover a subset of possible scenarios due to practical constraints [3]. The formal verification process encompasses several key roles, such as organizational coordination, task allocation, code development, property proving, analyzing Counter Examples (CEXs), debugging, coverage closure, and documentation preparation. These roles are crucial for managing the complexity and ensuring the thoroughness of the verification process. For instance, analyzing counterexamples involves identifying specific scenarios where the design might fail to meet its specifications, which is critical for debugging and refining the design. This highly intricate activity demands meticulous attention to detail, given its long development cycles and the critical nature of ensuring hardware functionality and reliability [4]. The field of Natural Language Processing (NLP) has undergone a significant transformation with the advent of Large Language Models (LLMs) [5].

Fifteen minutes after "hanging up" with Sesame's new "lifelike" AI, and I'm still freaked out. So-called "conversations" with AI don't do a lot for me, especially where text is concerned. With voice chats, such as the new options for Google Gemini and Microsoft's Copilot, all voice does is save some typing. While Google and Microsoft designed its assistants to be helpful, they're not especially personable -- or sometimes, they're just artificially cheery. Sesame's model, however, is a simple one: "We believe in a future where computers are lifelike," according to the company's mission statement.

The release of ChatGPT by OpenAI in November 2022 marked another significant milestone in the evolution of AI. ChatGPT, a large language model capable of generating human-like text, demonstrated the potential of AI to understand and generate natural language. This capability opened up new possibilities for AI applications, from customer service to content creation. The world responded to ChatGPT with a mix of awe and excitement, recognizing the potential of AI to transform how humans communicate and interact with technology to enhance our lives. Today, the rise of agentic AI -- systems capable of advanced reasoning and task execution -- is revolutionizing the way organizations operate.

OpenAI has unveiled its latest AI model, GPT-4.5, but the firm's boss says it is running out of hardware to power it. If ever-larger AI can no longer be run at scale, then are we looking at the end of the technology's rapid progress, and perhaps even the bursting of a bubble? There are certainly signs that things aren't going as planned within OpenAI. As recently as 12 February, CEO Sam Altman acknowledged on X that the company's product offering had created a confusing picture – at the…