Language Models over Canonical Byte-Pair Encodings

Modern language models represent probability distributions over character strings as distributions over (shorter) token strings derived via a deterministic tokenizer, such as byte-pair encoding. While this approach is highly effective at scaling up language models to large corpora, its current incarnations have a concerning property: the model assigns nonzero probability mass to an exponential number of $\it{noncanonical}$ token encodings of each character string -- these are token strings that decode to valid character strings but are impossible under the deterministic tokenizer (i.e., they will never be seen in any training corpus, no matter how large). This misallocation is both erroneous, as noncanonical strings never appear in training data, and wasteful, diverting probability mass away from plausible outputs. These are avoidable mistakes! In this work, we propose methods to enforce canonicality in token-level language models, ensuring that only canonical token strings are assigned positive probability. We present two approaches: (1) canonicality by conditioning, leveraging test-time inference strategies without additional training, and (2) canonicality by construction, a model parameterization that guarantees canonical outputs but requires training. We demonstrate that fixing canonicality mistakes improves the likelihood of held-out data for several models and corpora.

Well Begun is Half Done: Low-resource Preference Alignment by Weak-to-Strong Decoding

Large Language Models (LLMs) require alignment with human preferences to avoid generating offensive, false, or meaningless content. Recently, low-resource methods for LLM alignment have been popular, while still facing challenges in obtaining both high-quality and aligned content. Motivated by the observation that the difficulty of generating aligned responses is concentrated at the beginning of decoding, we propose a novel framework, Weak-to-Strong Decoding (WSD), to enhance the alignment ability of base models by the guidance of a small aligned model. The small model first drafts well-aligned beginnings, followed by the large base model to continue the rest, controlled by a well-designed auto-switch mechanism. We also collect a new dataset, GenerAlign, to fine-tune a small-sized Pilot-3B as the draft model, which effectively enhances different base models under the WSD framework to outperform all baseline methods, while avoiding degradation on downstream tasks, termed as the alignment tax. Extensive experiments are further conducted to examine the impact of different settings and time efficiency, as well as analyses on the intrinsic mechanisms of WSD in depth.

G1: Bootstrapping Perception and Reasoning Abilities of Vision-Language Model via Reinforcement Learning

Vision-Language Models (VLMs) excel in many direct multimodal tasks but struggle to translate this prowess into effective decision-making within interactive, visually rich environments like games. This ``knowing-doing'' gap significantly limits their potential as autonomous agents, as leading VLMs often performing badly in simple games. To address this, we introduce VLM-Gym, a curated reinforcement learning (RL) environment featuring diverse visual games with unified interfaces and adjustable, compositional difficulty, specifically designed for scalable multi-game parallel training. Leveraging VLM-Gym, we train G0 models using pure RL-driven self-evolution, which demonstrate emergent perception and reasoning patterns. To further mitigate challenges arising from game diversity, we develop G1 models. G1 incorporates a perception-enhanced cold start prior to RL fine-tuning. Our resulting G1 models consistently surpass their teacher across all games and outperform leading proprietary models like Claude-3.7-Sonnet-Thinking. Systematic analysis reveals an intriguing finding: perception and reasoning abilities mutually bootstrap each other throughout the RL training process. Source code including VLM-Gym and RL training are released at https://github.com/chenllliang/G1 to foster future research in advancing VLMs as capable interactive agents.

CellVerse: Do Large Language Models Really Understand Cell Biology?

Recent studies have demonstrated the feasibility of modeling single-cell data as natural languages and the potential of leveraging powerful large language models (LLMs) for understanding cell biology. However, a comprehensive evaluation of LLMs' performance on language-driven single-cell analysis tasks still remains unexplored. Motivated by this challenge, we introduce CellVerse, a unified language-centric question-answering benchmark that integrates four types of single-cell multi-omics data and encompasses three hierarchical levels of single-cell analysis tasks: cell type annotation (cell-level), drug response prediction (drug-level), and perturbation analysis (gene-level). Going beyond this, we systematically evaluate the performance across 14 open-source and closed-source LLMs ranging from 160M to 671B on CellVerse. Remarkably, the experimental results reveal: (1) Existing specialist models (C2S-Pythia) fail to make reasonable decisions across all sub-tasks within CellVerse, while generalist models such as Qwen, Llama, GPT, and DeepSeek family models exhibit preliminary understanding capabilities within the realm of cell biology. (2) The performance of current LLMs falls short of expectations and has substantial room for improvement. Notably, in the widely studied drug response prediction task, none of the evaluated LLMs demonstrate significant performance improvement over random guessing. CellVerse offers the first large-scale empirical demonstration that significant challenges still remain in applying LLMs to cell biology. By introducing CellVerse, we lay the foundation for advancing cell biology through natural languages and hope this paradigm could facilitate next-generation single-cell analysis.

Towards Artificial Intelligence Research Assistant for Expert-Involved Learning

Large Language Models (LLMs) and Large Multi-Modal Models (LMMs) have emerged as transformative tools in scientific research, yet their reliability and specific contributions to biomedical applications remain insufficiently characterized. In this study, we present \textbf{AR}tificial \textbf{I}ntelligence research assistant for \textbf{E}xpert-involved \textbf{L}earning (ARIEL), a multimodal dataset designed to benchmark and enhance two critical capabilities of LLMs and LMMs in biomedical research: summarizing extensive scientific texts and interpreting complex biomedical figures. To facilitate rigorous assessment, we create two open-source sets comprising biomedical articles and figures with designed questions. We systematically benchmark both open- and closed-source foundation models, incorporating expert-driven human evaluations conducted by doctoral-level experts. Furthermore, we improve model performance through targeted prompt engineering and fine-tuning strategies for summarizing research papers, and apply test-time computational scaling to enhance the reasoning capabilities of LMMs, achieving superior accuracy compared to human-expert corrections. We also explore the potential of using LMM Agents to generate scientific hypotheses from diverse multimodal inputs. Overall, our results delineate clear strengths and highlight significant limitations of current foundation models, providing actionable insights and guiding future advancements in deploying large-scale language and multi-modal models within biomedical research.

Deployment Optimization for XL-IRS Assisted Multi-User Communications

In this paper, we study the deployment optimization for an extremely large-scale intelligent reflecting surface (XL-IRS) assisted multi-user communication system, within which the channels between the XL-IRS and the BS (or user) are modeled by the near-field spherical wavefronts. To draw some valuable insights, we first consider the single-user case, where an alternating optimization (AO) based algorithm is devised to maximize the received signal-to-noise ratio (SNR) at the user. To address the high computational complexity issue incurred by the AO based algorithm, three approximate received SNR expressions are obtained to yield useful insights, corresponding to the upper bound, approximate expression, and closed-form. It is demonstrated that the XL-IRS ought to be positioned near the user (rather than the BS) to obtain a higher beamforming gain. Then, for the multi-user scenario, an efficient algorithm is proposed to obtain a high-quality XL-IRS placement solution by using the AO and successive convex approximation (SCA) techniques. Furthermore, the effective degree of freedom (DoF) of the BS-IRS channel is provided, which indicates that the additional effective DoF can be leveraged to improve multi-user spatial multiplexing. Last, numerical results confirm the existence of a trade-off between near-field beam-focusing gain and multiplexing gain.

Physical-Layer Security in Mixed Near-Field and Far-Field Communication Systems

Extremely large-scale arrays (XL-arrays) have emerged as a promising technology to improve the spectrum efficiency and spatial resolution of future wireless systems. Different from existing works that mostly considered physical layer security (PLS) in either the far-field or near-field, we consider in this paper a new and practical scenario, where legitimate users (Bobs) are located in the far-field of a base station (BS) while eavesdroppers (Eves) are located in the near-field for intercepting confidential information at short distance, referred to as the mixed near-field and far-field PLS. Specifically, we formulate an optimization problem to maximize the sum-secrecy-rate of all Bobs by optimizing the power allocation of the BS, subject to the constraint on the total BS transmit power. To shed useful insights, we first consider a one-Bob-one-Eve system and characterize the insecure-transmission region of the Bob in closed form. Interestingly, we show that the insecure-transmission region is significantly \emph{expanded} as compared to that in conventional far-field PLS systems, due to the energy-spread effect in the mixed-field scenario. Then, we further extend the analysis to a two-Bob-one-Eve system. It is revealed that as compared to the one-Bob system, the interferences from the other Bob can be effectively used to weaken the capability of Eve for intercepting signals of target Bobs, thus leading to enhanced secrecy rates. Furthermore, we propose an efficient algorithm to obtain a high-quality solution to the formulated non-convex problem by leveraging the successive convex approximation (SCA) technique. Finally, numerical results demonstrate that our proposed algorithm achieves a higher sum-secrecy-rate than the benchmark scheme where the power allocation is designed based on the (simplified) far-field channel model.

Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlo

A wide range of LM applications require generating text that conforms to syntactic or semantic constraints. Imposing such constraints can be naturally framed as probabilistic conditioning, but exact generation from the resulting distribution -- which can differ substantially from the LM's base distribution -- is generally intractable. In this work, we develop an architecture for controlled LM generation based on sequential Monte Carlo (SMC). Our SMC framework allows us to flexibly incorporate domain- and problem-specific constraints at inference time, and efficiently reallocate computational resources in light of new information during the course of generation. By comparing to a number of alternatives and ablations on four challenging domains -- Python code generation for data science, text-to-SQL, goal inference, and molecule synthesis -- we demonstrate that, with little overhead, our approach allows small open-source language models to outperform models over 8x larger, as well as closed-source, fine-tuned ones. In support of the probabilistic perspective, we show that these performance improvements are driven by better approximation to the posterior distribution. Our system builds on the framework of Lew et al. (2023) and integrates with its language model probabilistic programming language, giving users a simple, programmable way to apply SMC to a broad variety of controlled generation problems.

Multi-SWE-bench: A Multilingual Benchmark for Issue Resolving

The task of issue resolving is to modify a codebase to generate a patch that addresses a given issue. However, existing benchmarks, such as SWE-bench, focus almost exclusively on Python, making them insufficient for evaluating Large Language Models (LLMs) across diverse software ecosystems. To address this, we introduce a multilingual issue-resolving benchmark, called Multi-SWE-bench, covering Java, TypeScript, JavaScript, Go, Rust, C, and C++. It includes a total of 1,632 high-quality instances, which were carefully annotated from 2,456 candidates by 68 expert annotators, ensuring that the benchmark can provide an accurate and reliable evaluation. Based on Multi-SWE-bench, we evaluate a series of state-of-the-art models using three representative methods (Agentless, SWE-agent, and OpenHands) and present a comprehensive analysis with key empirical insights. In addition, we launch a Multi-SWE-RL open-source community, aimed at building large-scale reinforcement learning (RL) training datasets for issue-resolving tasks. As an initial contribution, we release a set of 4,723 well-structured instances spanning seven programming languages, laying a solid foundation for RL research in this domain. More importantly, we open-source our entire data production pipeline, along with detailed tutorials, encouraging the open-source community to continuously contribute and expand the dataset. We envision our Multi-SWE-bench and the ever-growing Multi-SWE-RL community as catalysts for advancing RL toward its full potential, bringing us one step closer to the dawn of AGI.

CodeCriticBench: A Holistic Code Critique Benchmark for Large Language Models

The critique capacity of Large Language Models (LLMs) is essential for reasoning abilities, which can provide necessary suggestions (e.g., detailed analysis and constructive feedback). Therefore, how to evaluate the critique capacity of LLMs has drawn great attention and several critique benchmarks have been proposed. However, existing critique benchmarks usually have the following limitations: (1). Focusing on diverse reasoning tasks in general domains and insufficient evaluation on code tasks (e.g., only covering code generation task), where the difficulty of queries is relatively easy (e.g., the code queries of CriticBench are from Humaneval and MBPP). (2). Lacking comprehensive evaluation from different dimensions. To address these limitations, we introduce a holistic code critique benchmark for LLMs called CodeCriticBench. Specifically, our CodeCriticBench includes two mainstream code tasks (i.e., code generation and code QA) with different difficulties. Besides, the evaluation protocols include basic critique evaluation and advanced critique evaluation for different characteristics, where fine-grained evaluation checklists are well-designed for advanced settings. Finally, we conduct extensive experimental results of existing LLMs, which show the effectiveness of CodeCriticBench.