-- Speculative decoding has emerged as an effective approach for accelerating autoregressive inference by parallelizing token generation through a draft - then - verify paradigm. However, existing methods rely on static drafting lengths and rigid verification cri teria, limiting their adaptability across varying model uncertainties and input complexities. This paper proposes an information - theoretic framework for speculative decoding based on confidence - modulated drafting. By leveraging entropy and margin - based uncertainty measures over the drafter's output distribution, the proposed method dynamically adjusts the number of speculatively generated tokens at each iteration. This adaptive mechanism reduces rollback frequency, improves resource utilization, an d maintains output fidelity. Additionally, the verification process is modulated using the same confidence signals, enabling more flexible acceptance of drafted tokens without sacrificing generation quality. Experiments on machine translation and summariza tion tasks demonstrate significant speedups over standard speculative decoding while preserving or improving BLEU and ROUGE scores. The proposed approach offers a principled, plug - in method for efficient and robust decoding in large language models under v arying conditions of uncertainty. Keywords -- Speculative Decoding, Autoregressive Models, Confidence Estimation, Adaptive Inference, Entropy - Based Drafting, Sequence Generation, Large Language Models, Large Language Models (LLMs), Information - Theoretic Decoding. The task of sequence generation lies at the heart of numerous applications in natural language processing, including machine translation, text summarization, dialogue generation, and code synthesis. In the overwhelming majority of these applications, autor egressive (AR) decoding remains the dominant paradigm for generating sequences from a probabilistic language model [1 - 2] . Autoregressive models, particularly those based on the Transformer architecture, operate by predicting each token conditioned on the e ntire history of previously generated tokens. This left - to - right decoding strategy, though optimal in terms of likelihood estimation, suffers from a fundamental limitation: the inherently sequential nature of generation prohibits efficient parallelization, severely hindering inference throughput, especially in latency - sensitive deployment scenarios.

Partitioning large machine learning models across distributed accelerator systems is a complex process, requiring a series of interdependent decisions that are further complicated by internal sharding ambiguities. Consequently, existing auto-partitioners often suffer from out-of-memory errors or are prohibitively slow when exploring the exponentially large space of possible partitionings. To mitigate this, they artificially restrict the search space, but this approach frequently yields infeasible solutions that violate device memory constraints or lead to sub-optimal performance. We propose a system that combines a novel static compiler analysis with a Monte Carlo Tree Search. Our analysis constructs an efficient decision space by identifying (i) tensor dimensions requiring identical sharding, and (ii) partitioning "conflicts" that require resolution. Our system significantly outperforms state-of-the-art industrial methods across diverse hardware platforms and model architectures, discovering previously unknown, superior solutions, and the process is fully automated even for complex and large models.

We present the preliminary rankings of machine translation (MT) systems submitted to the WMT25 General Machine Translation Shared Task, as determined by automatic evaluation metrics. Because these rankings are derived from automatic evaluation, they may exhibit a bias toward systems that employ re-ranking techniques, such as Quality Estimation or Minimum Bayes Risk decoding. The official WMT25 ranking will be based on human evaluation, which is more reliable and will supersede these results. The official WMT25 ranking will be based on human evaluation, which is more reliable and will supersede these results. The purpose of releasing these findings now is to assist task participants with their system description papers; not to provide final findings.

As models grow larger and training them becomes expensive, it becomes increasingly important to scale training recipes not just to larger models and more data, but to do so in a compute-optimal manner that extracts maximal performance per unit of compute. While such scaling has been well studied for language modeling, reinforcement learning (RL) has received less attention in this regard. In this paper, we investigate compute scaling for online, value-based deep RL. These methods present two primary axes for compute allocation: model capacity and the update-to-data (UTD) ratio. Given a fixed compute budget, we ask: how should resources be partitioned across these axes to maximize sample efficiency? Our analysis reveals a nuanced interplay between model size, batch size, and UTD. In particular, we identify a phenomenon we call TD-overfitting: increasing the batch quickly harms Q-function accuracy for small models, but this effect is absent in large models, enabling effective use of large batch size at scale. We provide a mental model for understanding this phenomenon and build guidelines for choosing batch size and UTD to optimize compute usage. Our findings provide a grounded starting point for compute-optimal scaling in deep RL, mirroring studies in supervised learning but adapted to TD learning.

Large language models (LLMs) have demonstrated potential in educational applications, yet their capacity to accurately assess the cognitive alignment of reading materials with students' developmental stages remains insufficiently explored. This gap is particularly critical given the foundational educational principle of the Zone of Proximal Development (ZPD), which emphasizes the need to match learning resources with Students' Cognitive Abilities (SCA). Despite the importance of this alignment, there is a notable absence of comprehensive studies investigating LLMs' ability to evaluate reading comprehension difficulty across different student age groups, especially in the context of Chinese language education. To fill this gap, we introduce ZPD-SCA, a novel benchmark specifically designed to assess stage-level Chinese reading comprehension difficulty. The benchmark is annotated by 60 Special Grade teachers, a group that represents the top 0.15% of all in-service teachers nationwide. Experimental results reveal that LLMs perform poorly in zero-shot learning scenarios, with Qwen-max and GLM even falling below the probability of random guessing. When provided with in-context examples, LLMs performance improves substantially, with some models achieving nearly double the accuracy of their zero-shot baselines. These results reveal that LLMs possess emerging abilities to assess reading difficulty, while also exposing limitations in their current training for educationally aligned judgment. Notably, even the best-performing models display systematic directional biases, suggesting difficulties in accurately aligning material difficulty with SCA. Furthermore, significant variations in model performance across different genres underscore the complexity of task. We envision that ZPD-SCA can provide a foundation for evaluating and improving LLMs in cognitively aligned educational applications.

As researchers increasingly adopt LLMs as writing assistants, generating high-quality research paper introductions remains both challenging and essential. We introduce Scientific Introduction Generation (SciIG), a task that evaluates LLMs' ability to produce coherent introductions from titles, abstracts, and related works. Curating new datasets from NAACL 2025 and ICLR 2025 papers, we assess five state-of-the-art models, including both open-source (DeepSeek-v3, Gemma-3-12B, LLaMA 4-Maverick, MistralAI Small 3.1) and closed-source GPT-4o systems, across multiple dimensions: lexical overlap, semantic similarity, content coverage, faithfulness, consistency, citation correctness, and narrative quality. Our comprehensive framework combines automated metrics with LLM-as-a-judge evaluations. Results demonstrate LLaMA-4 Maverick's superior performance on most metrics, particularly in semantic similarity and faithfulness. Moreover, three-shot prompting consistently outperforms fewer-shot approaches. These findings provide practical insights into developing effective research writing assistants and set realistic expectations for LLM-assisted academic writing. To foster reproducibility and future research, we will publicly release all code and datasets.

Live cell culture is crucial in biomedical studies for analyzing cell properties and dynamics in vitro. This study focuses on segmenting unstained live cells imaged with bright-field microscopy. While many segmentation approaches exist for microscopic images, none consistently address the challenges of bright-field live-cell imaging with high throughput, where temporal phenotype changes, low contrast, noise, and motion-induced blur from cellular movement remain major obstacles. We developed a low-cost CNN-based pipeline incorporating comparative analysis of frozen encoders within a unified U-Net architecture enhanced with attention mechanisms, instance-aware systems, adaptive loss functions, hard instance retraining, dynamic learning rates, progressive mechanisms to mitigate overfitting, and an ensemble technique. The model was validated on a public dataset featuring diverse live cell variants, showing consistent competitiveness with state-of-the-art methods, achieving 93% test accuracy and an average F1-score of 89% (std. 0.07) on low-contrast, noisy, and blurry images. Notably, the model was trained primarily on bright-field images with limited exposure to phase- contrast microscopy (<20%), yet it generalized effectively to the phase-contrast LIVECell dataset, demonstrating modality, robustness and strong performance. This highlights its potential for real- world laboratory deployment across imaging conditions. The model requires minimal compute power and is adaptable using basic deep learning setups such as Google Colab, making it practical for training on other cell variants. Our pipeline outperforms existing methods in robustness and precision for bright-field microscopy segmentation. The code and dataset are available for reproducibility 1.

Modern GPUs are equipped with large amounts of high-bandwidth memory, enabling them to support mini-batch sizes of up to tens of thousands of training samples. However, most existing optimizers struggle to perform effectively at such a large batch size. As batch size increases, gradient noise decreases due to averaging over many samples, limiting the ability of first-order methods to escape sharp or suboptimal minima and reach the global minimum. Meanwhile, second-order methods like the natural gradient with Kronecker-Factored Approximate Curvature (KFAC) often require excessively high damping to remain stable at large batch sizes. This high damping effectively "washes out" the curvature information that gives these methods their advantage, reducing their performance to that of simple gradient descent. In this paper, we introduce Fisher-Orthogonal Projection (FOP), a novel technique that restores the effectiveness of the second-order method at very large batch sizes, enabling scalable training with improved generalization and faster convergence. FOP constructs a variance-aware update direction by leveraging gradients from two sub-batches, enhancing the average gradient with a component of the gradient difference that is orthogonal to the average under the Fisher-metric. Through extensive benchmarks, we show that FOP accelerates convergence by 1 .2-1. 3 over KFAC and 1 .

Large Language Models have shown growing ability to generate fluent and coherent texts that are highly similar to the writing style of humans. Current detectors for Machine-Generated Text (MGT) perform well when they are trained and tested in the same domain but generalize poorly to unseen domains, due to domain shift between data from different sources. In this work, we propose MGT-Prism, an MGT detection method from the perspective of the frequency domain for better domain generalization. Our key insight stems from analyzing text representations in the frequency domain, where we observe consistent spectral patterns across diverse domains, while significant discrepancies in magnitude emerge between MGT and human-written texts (HWTs). The observation initiates the design of a low frequency domain filtering module for filtering out the document-level features that are sensitive to domain shift, and a dynamic spectrum alignment strategy to extract the task-specific and domain-invariant features for improving the detector's performance in domain generalization. Extensive experiments demonstrate that MGT-Prism outperforms state-of-the-art baselines by an average of 0.90% in accuracy and 0.92% in F1 score on 11 test datasets across three domain-generalization scenarios.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable performance in complex multimodal tasks. While MLLMs excel at visual perception and reasoning in third-person and egocentric videos, they are prone to hallucinations, generating coherent yet inaccurate responses. We present EgoIllusion, a first benchmark to evaluate MLLM hallucinations in egocentric videos. EgoIllusion comprises 1,400 videos paired with 8,000 human-annotated open and closed-ended questions designed to trigger hallucinations in both visual and auditory cues in egocentric videos. Evaluations across ten MLLMs reveal significant challenges, including powerful models like GPT-4o and Gemini, achieving only 59% accuracy. EgoIllusion lays the foundation in developing robust benchmarks to evaluate the effectiveness of MLLMs and spurs the development of better egocentric MLLMs with reduced hallucination rates. Our benchmark will be open-sourced for reproducibility.