Large language models (LLMs) have recently demonstrated remarkable capabilities in machine translation (MT). However, most advanced MT-specific LLMs heavily rely on external supervision signals during training, such as human-annotated reference data or trained reward models (RMs), which are often expensive to obtain and challenging to scale. To overcome this limitation, we propose a Simple Self-Rewarding (SSR) Reinforcement Learning (RL) framework for MT that is reference-free, fully online, and relies solely on self-judging rewards. Training with SSR using 13K monolingual examples and Qwen-2.5-7B as the backbone, our model SSR-Zero-7B outperforms existing MT-specific LLMs, e.g., TowerInstruct-13B and GemmaX-28-9B, as well as larger general LLMs like Qwen2.5-32B-Instruct in English $\leftrightarrow$ Chinese translation tasks from WMT23, WMT24, and Flores200 benchmarks. Furthermore, by augmenting SSR with external supervision from COMET, our strongest model, SSR-X-Zero-7B, achieves state-of-the-art performance in English $\leftrightarrow$ Chinese translation, surpassing all existing open-source models under 72B parameters and even outperforming closed-source models, e.g., GPT-4o and Gemini 1.5 Pro. Our analysis highlights the effectiveness of the self-rewarding mechanism compared to the external LLM-as-a-judge approach in MT and demonstrates its complementary benefits when combined with trained RMs. Our findings provide valuable insight into the potential of self-improving RL methods. We have publicly released our code, data and models.

The increasing demand for generative AI as Large Language Models (LLMs) services has driven the need for specialized hardware architectures that optimize computational efficiency and energy consumption. This paper evaluates the performance of the Tenstorrent Grayskull e75 RISC-V accelerator for basic linear algebra kernels at reduced numerical precision, a fundamental operation in LLM computations. We present a detailed characterization of Grayskull's execution model, grid size, matrix dimensions, data formats, and numerical precision impact on computational efficiency. Furthermore, we compare Grayskull's performance against state-of-the-art architectures with tensor acceleration, including Intel Sapphire Rapids processors and two NVIDIA GPUs (V100 and A100). Whilst NVIDIA GPUs dominate raw performance, Grayskull demonstrates a competitive trade-off between power consumption and computational throughput, reaching a peak of 1.55 TFLOPs/Watt with BF16.

Urban causal research is essential for understanding the complex, dynamic processes that shape cities and for informing evidence-based policies. However, current practices are often constrained by inefficient and biased hypothesis formulation, challenges in integrating multimodal data, and fragile experimental methodologies. Imagine a system that automatically estimates the causal impact of congestion pricing on commute times by income group or measures how new green spaces affect asthma rates across neighborhoods using satellite imagery and health reports, and then generates comprehensive, policy-ready outputs, including causal estimates, subgroup analyses, and actionable recommendations. In this Perspective, we propose UrbanCIA, an LLM-driven conceptual framework composed of four distinct modular agents responsible for hypothesis generation, data engineering, experiment design and execution, and results interpretation with policy insights. We begin by examining the current landscape of urban causal research through a structured taxonomy of research topics, data sources, and methodological approaches, revealing systemic limitations across the workflow. Next, we introduce the design principles and technological roadmap for the four modules in the proposed framework. We also propose evaluation criteria to assess the rigor and transparency of these AI-augmented processes. Finally, we reflect on the broader implications for human-AI collaboration, equity, and accountability. We call for a new research agenda that embraces LLM-driven tools as catalysts for more scalable, reproducible, and inclusive urban research.

Prompting has become a dominant paradigm for adapting large language models (LLMs). While discrete (textual) prompts are widely used for their interpretability, soft (parameter) prompts have recently gained traction in APIs. This is because they can encode information from more training samples while minimizing the user's token usage, leaving more space in the context window for task-specific input. However, soft prompts are tightly coupled to the LLM they are tuned on, limiting their generalization to other LLMs. This constraint is particularly problematic for efficiency and privacy: (1) tuning prompts on each LLM incurs high computational costs, especially as LLMs continue to grow in size. Additionally, (2) when the LLM is hosted externally, soft prompt tuning often requires sharing private data with the LLM provider. For instance, this is the case with the NVIDIA NeMo API. To address these issues, we propose POST (Privacy Of Soft prompt Transfer), a framework that enables private tuning of soft prompts on a small model and subsequently transfers these prompts to a larger LLM. POST uses knowledge distillation to derive a small model directly from the large LLM to improve prompt transferability, tunes the soft prompt locally, optionally with differential privacy guarantees, and transfers it back to the larger LLM using a small public dataset. Our experiments show that POST reduces computational costs, preserves privacy, and effectively transfers high-utility soft prompts.

In this work, we propose JETHICS, a Japanese dataset for evaluating ethics understanding of AI models. JETHICS contains 78K examples and is built by following the construction methods of the existing English ETHICS dataset. It includes four categories based normative theories and concepts from ethics and political philosophy; and one representing commonsense morality. Our evaluation experiments on non-proprietary large language models (LLMs) and on GPT-4o reveal that even GPT-4o achieves only an average score of about 0.7, while the best-performing Japanese LLM attains around 0.5, indicating a relatively large room for improvement in current LLMs.

Recent research in retrieval-augmented generation (RAG) has concentrated on retrieving useful information from candidate documents. However, numerous methodologies frequently neglect the calibration capabilities of large language models (LLMs), which capitalize on their robust in-context reasoning prowess. This work illustrates that providing LLMs with specific cues substantially improves their calibration efficacy, especially in multi-round calibrations. We present a new SGIC: Self-Guided Iterative Calibration Framework that employs uncertainty scores as a tool. Initially, this framework calculates uncertainty scores to determine both the relevance of each document to the query and the confidence level in the responses produced by the LLMs. Subsequently, it reevaluates these scores iteratively, amalgamating them with prior responses to refine calibration. Furthermore, we introduce an innovative approach for constructing an iterative self-calibration training set, which optimizes LLMs to efficiently harness uncertainty scores for capturing critical information and enhancing response accuracy. Our proposed framework significantly improves performance on both closed-source and open-weight LLMs.

Imagine unlocking the power of the mind to communicate, create, and even interact with the world around us. Recent breakthroughs in Artificial Intelligence (AI), especially in how machines "see" and "understand" language, are now fueling exciting progress in decoding brain signals from scalp electroencephalography (EEG). Prima facie, this opens the door to revolutionary brain-computer interfaces (BCIs) designed for real life, moving beyond traditional uses to envision Brain-to-Speech, Brain-to-Image, and even a Brain-to-Internet of Things (BCIoT). However, the journey is not as straightforward as it was for Computer Vision (CV) and Natural Language Processing (NLP). Applying AI to real-world EEG-based BCIs, particularly in building powerful foundational models, presents unique and intricate hurdles that could affect their reliability. Here, we unfold a guided exploration of this dynamic and rapidly evolving research area. Rather than barely outlining a map of current endeavors and results, the goal is to provide a principled navigation of this hot and cutting-edge research landscape. We consider the basic paradigms that emerge from a causal perspective and the attendant challenges presented to AI-based models. Looking ahead, we then discuss promising research avenues that could overcome today's technological, methodological, and ethical limitations. Our aim is to lay out a clear roadmap for creating truly practical and effective EEG-based BCI solutions that can thrive in everyday environments.

Human decision-making belongs to the foundation of our society and civilization, but we are on the verge of a future where much of it will be delegated to artificial intelligence. The arrival of Large Language Models (LLMs) has transformed the nature and scope of AI-supported decision-making; however, the process by which they learn to make decisions, compared to humans, remains poorly understood. In this study, we examined the decision-making behavior of five leading LLMs across three core dimensions of real-world decision-making: uncertainty, risk, and set-shifting. Using three well-established experimental psychology tasks designed to probe these dimensions, we benchmarked LLMs against 360 newly recruited human participants. Across all tasks, LLMs often outperformed humans, approaching near-optimal performance. Moreover, the processes underlying their decisions diverged fundamentally from those of humans. On the one hand, our finding demonstrates the ability of LLMs to manage uncertainty, calibrate risk, and adapt to changes. On the other hand, this disparity highlights the risks of relying on them as substitutes for human judgment, calling for further inquiry.

Language is not only a tool for communication but also a medium for human cognition and reasoning. If, as linguistic relativity suggests, the structure of language shapes cognitive patterns, then large language models (LLMs) trained on human language may also internalize the habitual logical structures embedded in different languages. To examine this hypothesis, we introduce BICAUSE, a structured bilingual dataset for causal reasoning, which includes semantically aligned Chinese and English samples in both forward and reversed causal forms. Our study reveals three key findings: (1) LLMs exhibit typologically aligned attention patterns, focusing more on causes and sentence-initial connectives in Chinese, while showing a more balanced distribution in English. (2) Models internalize language-specific preferences for causal word order and often rigidly apply them to atypical inputs, leading to degraded performance, especially in Chinese. (3) When causal reasoning succeeds, model representations converge toward semantically aligned abstractions across languages, indicating a shared understanding beyond surface form. Overall, these results suggest that LLMs not only mimic surface linguistic forms but also internalize the reasoning biases shaped by language. Rooted in cognitive linguistic theory, this phenomenon is for the first time empirically verified through structural analysis of model internals.

Recent reinforcement learning approaches, such as outcome-supervised GRPO, have advanced Chain-of-Thought reasoning in large language models (LLMs), yet their adaptation to multimodal LLMs (MLLMs) is unexplored. To address the lack of rigorous evaluation for MLLM post-training methods, we introduce SEED-Bench-R1, a benchmark with complex real-world videos requiring balanced perception and reasoning. It offers a large training set and evaluates generalization across three escalating challenges: in-distribution, cross-environment, and cross-environment-task scenarios. Using SEED-Bench-R1, we find that standard GRPO, while improving answer accuracy, often reduces logical coherence between reasoning steps and answers, with only a 57.9% consistency rate. This stems from reward signals focusing solely on final answers, encouraging shortcuts, and strict KL penalties limiting exploration.To address this, we propose GRPO-CARE, a consistency-aware RL framework optimizing both answer correctness and reasoning coherence without explicit supervision. GRPO-CARE introduces a two-tiered reward: (1) a base reward for answer correctness, and (2) an adaptive consistency bonus, computed by comparing the model's reasoning-to-answer likelihood (via a slowly-evolving reference model) against group peers.This dual mechanism amplifies rewards for reasoning paths that are both correct and logically consistent. Replacing KL penalties with this adaptive bonus, GRPO-CARE outperforms standard GRPO on SEED-Bench-R1, achieving a 6.7% performance gain on the hardest evaluation level and a 24.5% improvement in consistency. It also shows strong transferability, improving model performance across diverse video understanding benchmarks. Our work contributes a systematically designed benchmark and a generalizable post-training framework, advancing the development of more interpretable and robust MLLMs.