Yanjun Chen

Yunfan Jiang, Agrim Gupta, Zichen Zhang et al. · stanford

Prompt-based learning has emerged as a successful paradigm in natural language processing, where a single general-purpose language model can be instructed to perform any task specified by input prompts. Yet task specification in robotics comes in various forms, such as imitating one-shot demonstrations, following language instructions, and reaching visual goals. They are often considered different tasks and tackled by specialized models. We show that a wide spectrum of robot manipulation tasks can be expressed with multimodal prompts, interleaving textual and visual tokens. Accordingly, we develop a new simulation benchmark that consists of thousands of procedurally-generated tabletop tasks with multimodal prompts, 600K+ expert trajectories for imitation learning, and a four-level evaluation protocol for systematic generalization. We design a transformer-based robot agent, VIMA, that processes these prompts and outputs motor actions autoregressively. VIMA features a recipe that achieves strong model scalability and data efficiency. It outperforms alternative designs in the hardest zero-shot generalization setting by up to $2.9\times$ task success rate given the same training data. With $10\times$ less training data, VIMA still performs $2.7\times$ better than the best competing variant. Code and video demos are available at https://vimalabs.github.io/

Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri et al. · allen-ai, berkeley

Modern artificial intelligence (AI) systems are powered by foundation models. This paper presents a new set of foundation models, called Llama 3. It is a herd of language models that natively support multilinguality, coding, reasoning, and tool usage. Our largest model is a dense Transformer with 405B parameters and a context window of up to 128K tokens. This paper presents an extensive empirical evaluation of Llama 3. We find that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development.

Tianqi Xu, Linyao Chen, Dai-Jie Wu et al.

The development of autonomous agents increasingly relies on Multimodal Language Models (MLMs) to perform tasks described in natural language with GUI environments, such as websites, desktop computers, or mobile phones. Existing benchmarks for MLM agents in interactive environments are limited by their focus on a single environment, lack of detailed and generalized evaluation methods, and the complexities of constructing tasks and evaluators. To overcome these limitations, we introduce Crab, the first agent benchmark framework designed to support cross-environment tasks, incorporating a graph-based fine-grained evaluation method and an efficient mechanism for task and evaluator construction. Our framework supports multiple devices and can be easily extended to any environment with a Python interface. Leveraging Crab, we developed a cross-platform Crab Benchmark-v0 comprising 120 tasks in computer desktop and mobile phone environments. We evaluated four advanced MLMs using different single and multi-agent system configurations on this benchmark. The experimental results demonstrate that the single agent with GPT-4o achieves the best completion ratio of 38.01%. All framework code, agent code, and task datasets are publicly available at https://github.com/camel-ai/crab.

Yanjun Chen, Yirong Sun, Hanlin Wang et al.

Removing an agent from a cooperative team to measure its contribution seems natural, yet in multi-agent LLM systems this evaluation distorts the result it claims to measure. This failure is not isolated: learned critics, trajectory-level baselines, and agent-removal counterfactuals all inherit from standard multi-agent reinforcement learning a premise that exact counterfactual evaluation requires privileged environment access, and therefore approximate. In cooperative LLM systems, this premise is false. Interaction histories are deterministic functions of observable text with no hidden state, so any decision point can be restored exactly, making direct causal measurement possible without parametric approximation. C3 exploits this property by fixing the complete history at each decision point, sampling alternative actions under a frozen behavior policy, and computing unbiased per-decision advantages through a parameter-free leave-one-out baseline. Across six benchmarks spanning math reasoning and code generation, two model families, and two multi-agent topologies, C3 consistently outperforms all baselines; a controlled decomposition confirms gains originate from credit quality, not architecture, while checkpoint restoration reduces training token consumption. The exact solution proves simpler, cheaper, and more effective than all approximate alternatives. The same structural property that enables exact credit also enables exact verification: three independently computable diagnostics, credit fidelity, within-group variance, and inter-agent influence, constitute the first method-agnostic auditing tool for multi-agent LLM credit assignment. Our code is available at https://github.com/EIT-EAST-Lab/C3

Hao Zhou, Chongyang Zhang, Yanjun Chen et al.

Temporal grounding aims to retrieve moments of the described event within an untrimmed video by a language query. Typically, existing methods assume annotations are precise and unique, yet one query may describe multiple moments in many cases. Hence, simply taking it as a one-vs-one mapping task and striving to match single-label annotations will inevitably introduce false negatives during optimization. In this study, we reformulate this task as a one-vs-many optimization problem under the condition of single positive labels. The unlabeled moments are considered unobserved rather than negative, and we explore mining potential positive moments to assist in multiple moment retrieval. In this setting, we propose a novel Diverse Temporal Grounding framework, termed DTG-SPL, which mainly consists of a positive moment estimation (PME) module and a diverse moment regression (DMR) module. PME leverages semantic reconstruction information and an expected positive regularization to uncover potential positive moments in an online fashion. Under the supervision of these pseudo positives, DMR is able to localize diverse moments in parallel that meet different users. The entire framework allows for end-to-end optimization as well as fast inference. Extensive experiments on Charades-STA and ActivityNet Captions show that our method achieves superior performance in terms of both single-label and multi-label metrics.

Tian Dong, Yanjun Chen, Shoufeng Zhang et al.

AI infra has become a shared execution layer for model training, deployment, and agent orchestration. Because many projects reimplement similar model-centric workflows, a vulnerability disclosed in one repository can recur as a variant in another repository with a related design. Yet the prevalence and detectability of these variants remain poorly understood. This paper presents a measurement study of vulnerability variants in AI infra. Analyzing 688 GitHub repositories and 251 publicly disclosed vulnerabilities, we find that AI infra projects frequently share overlapping functionality and recurrent vulnerable patterns, creating a concrete basis for cross-repository variants. Building on this finding, we study how to automatically identify such variants from known disclosures. We propose INFRASCOPE, a reference-driven multi-agent framework that extracts transferable vulnerability semantics from known cases and uses them to locate and validate variants in new repositories. Evaluating INFRASCOPE on 20 real-world AI infra repositories, we uncover over 20 vulnerabilities, including 11 acknowledged cases and 4 cases that have been assigned CVEs so far.

Xinghao Chen, Zhijing Sun, Wenjin Guo et al.

Large Language Models (LLMs) excel in reasoning tasks through Chain-of-Thought (CoT) prompting. However, CoT prompting greatly increases computational demands, which has prompted growing interest in distilling CoT capabilities into Small Language Models (SLMs). This study systematically examines the factors influencing CoT distillation, including the choice of granularity, format and teacher model. Through experiments involving four teacher models and seven student models across seven mathematical and commonsense reasoning datasets, we uncover three key findings: (1) Unlike LLMs, SLMs exhibit a non-monotonic relationship with granularity, with stronger models benefiting from finer-grained reasoning and weaker models performing better with simpler CoT supervision; (2) CoT format significantly impacts LLMs but has minimal effect on SLMs, likely due to their reliance on supervised fine-tuning rather than pretraining preferences; (3) Stronger teacher models do NOT always produce better student models, as diversity and complexity in CoT supervision can outweigh accuracy alone. These findings emphasize the need to tailor CoT strategies to specific student model, offering actionable insights for optimizing CoT distillation in SLMs. The code and datasets are available at https://github.com/EIT-NLP/Distilling-CoT-Reasoning.

Yirong Sun, Yanjun Chen, Xin Qiu et al.

Large Audio Language Models (LALMs) excel at semantic and paralinguistic tasks, yet their ability to perceive the fundamental physical attributes of audio such as pitch, loudness, and spatial location remains under-explored. To bridge this gap, we introduce SonicBench, a psychophysically grounded benchmark that systematically evaluates 12 core physical attributes across five perceptual dimensions. Unlike previous datasets, SonicBench uses a controllable generation toolbox to construct stimuli for two complementary paradigms: recognition (absolute judgment) and comparison (relative judgment). This design allows us to probe not only sensory precision but also relational reasoning capabilities, a domain where humans typically exhibit greater proficiency. Our evaluation reveals a substantial deficiency in LALMs' foundational auditory understanding; most models perform near random guessing and, contrary to human patterns, fail to show the expected advantage on comparison tasks. Furthermore, explicit reasoning yields minimal gains. However, our linear probing analysis demonstrates crucially that frozen audio encoders do successfully capture these physical cues (accuracy at least 60%), suggesting that the primary bottleneck lies in the alignment and decoding stages, where models fail to leverage the sensory signals they have already captured.

Xinghao Chen, Anhao Zhao, Heming Xia et al.

Large Language Models (LLMs) have shown impressive performance on complex tasks through Chain-of-Thought (CoT) reasoning. However, conventional CoT relies on explicitly verbalized intermediate steps, which constrains its broader applicability, particularly in abstract reasoning tasks beyond language. To address this, there has been growing research interest in \textit{latent CoT reasoning}, where the reasoning process is embedded within latent spaces. By decoupling reasoning from explicit language generation, latent CoT offers the promise of richer cognitive representations and facilitates more flexible, faster inference. This paper aims to present a comprehensive overview of this emerging paradigm and establish a systematic taxonomy. We analyze recent advances in methods, categorizing them from token-wise horizontal approaches to layer-wise vertical strategies. We then provide in-depth discussions of these methods, highlighting their design principles, applications, and remaining challenges. We hope that our survey provides a structured foundation for advancing this promising direction in LLM reasoning. The relevant papers will be regularly updated at https://github.com/EIT-NLP/Awesome-Latent-CoT.

Yirong Sun, Dawei Zhu, Yanjun Chen et al.

Large language models (LLMs) have excelled in various NLP tasks, including machine translation (MT), yet most studies focus on sentence-level translation. This work investigates the inherent capability of instruction-tuned LLMs for document-level translation (docMT). Unlike prior approaches that require specialized techniques, we evaluate LLMs by directly prompting them to translate entire documents in a single pass. Our results show that this method improves translation quality compared to translating sentences separately, even without document-level fine-tuning. However, this advantage is not reflected in BLEU scores, which often favor sentence-based translations. We propose using the LLM-as-a-judge paradigm for evaluation, where GPT-4 is used to assess document coherence, accuracy, and fluency in a more nuanced way than n-gram-based metrics. Overall, our work demonstrates that instruction-tuned LLMs can effectively leverage document context for translation. However, we caution against using BLEU scores for evaluating docMT, as they often provide misleading outcomes, failing to capture the quality of document-level translation. Code and the outputs from GPT4-as-a-judge are available at https://github.com/EIT-NLP/BLEUless_DocMT

Yanjun Chen, Yirong Sun, Xinghao Chen et al.

Chain-of-Thought (CoT) reasoning has proven effective in natural language tasks but remains underexplored in multimodal alignment. This study investigates its integration into 3D vision-language learning by embedding structured reasoning into alignment training. We introduce the 3D-CoT Benchmark, a dataset with hierarchical CoT annotations covering shape recognition, functional inference, and causal reasoning. Through controlled experiments, we compare CoT-structured and standard textual annotations across large reasoning models (LRMs) and large language models (LLMs). Our evaluation employs a dual-layer framework assessing both intermediate reasoning and final inference quality. Extensive experiments demonstrate that CoT significantly improves 3D semantic grounding, with LRMs leveraging CoT more effectively than LLMs. Furthermore, we highlight that annotation structure influences performance-explicit reasoning markers aid LLMs, while unmarked CoT better aligns with LRM inference patterns. Our analyses suggest that CoT is crucial for enhancing multimodal reasoning, with implications beyond 3D tasks. The dataset will be publicly available at https://huggingface.co/datasets/Battam/3D-CoT

Yunuo Liu, Dawei Zhu, Zena Al-Khalili et al.

We present PricingLogic, the first benchmark that probes whether Large Language Models(LLMs) can reliably automate tourism-related prices when multiple, overlapping fare rules apply. Travel agencies are eager to offload this error-prone task onto AI systems; however, deploying LLMs without verified reliability could result in significant financial losses and erode customer trust. PricingLogic comprises 300 natural-language questions based on booking requests derived from 42 real-world pricing policies, spanning two levels of difficulty: (i) basic customer-type pricing and (ii)bundled-tour calculations involving interacting discounts. Evaluations of a line of LLMs reveal a steep performance drop on the harder tier,exposing systematic failures in rule interpretation and arithmetic reasoning.These results highlight that, despite their general capabilities, today's LLMs remain unreliable in revenue-critical applications without further safeguards or domain adaptation. Our code and dataset are available at https://github.com/EIT-NLP/PricingLogic.

Yirong Sun, Yizhong Geng, Peidong Wei et al.

The development of Large Speech-Language Models (LSLMs) has been slowed by fragmented architectures and a lack of transparency, hindering the systematic comparison and reproducibility of research. Unlike in the vision-language domain, the LSLM field suffers from the common practice of releasing model weights without their corresponding training data and configurations. To address these critical gaps, we introduce LLaSO, the first fully open, end-to-end framework for large-scale speech-language modeling. LLaSO provides the community with three essential resources: (1) LLaSO-Align, a 12M-instance speech-text alignment corpus; (2) LLaSO-Instruct, a 13.5M-instance multi-task instruction-tuning dataset; and (3) LLaSO-Eval, a reproducible benchmark for standardized evaluation. To validate our framework, we build and release LLaSO-Base, a 3.8B-parameter reference model trained exclusively on our public data. It achieves a normalized score of 0.72, establishing a strong, reproducible baseline that surpasses comparable models. Our analysis reveals that while broader training coverage enhances performance, significant generalization gaps persist on unseen tasks, particularly in pure audio scenarios. By releasing the complete stack of data, benchmarks, and models, LLaSO establishes a foundational open standard to unify research efforts and accelerate community-driven progress in LSLMs. We release the code, dataset, pretrained models, and results in https://github.com/EIT-NLP/LLaSO.

Jun Xue, Junze Wang, Xinming Zhang et al.

Scaling Maximum Entropy Reinforcement Learning (RL) to high-dimensional humanoid control remains a formidable challenge, as the ``curse of dimensionality'' induces severe exploration inefficiency and training instability in expansive action spaces. Consequently, recent high-throughput paradigms have largely converged on deterministic policy gradients combined with massive parallel simulation. We challenge this compromise with FastDSAC, a framework that effectively unlocks the potential of maximum entropy stochastic policies for complex continuous control. We introduce Dimension-wise Entropy Modulation (DEM) to dynamically redistribute the exploration budget and enforce diversity, alongside a continuous distributional critic tailored to ensure value fidelity and mitigate high-dimensional value overestimation. Extensive evaluations on HumanoidBench and other continuous control tasks demonstrate that rigorously designed stochastic policies can consistently match or outperform deterministic baselines, achieving notable gains of 180\% and 400\% on the challenging \textit{Basketball} and \textit{Balance Hard} tasks.

Yirong Sun, Yanjun Chen

We propose PRISM, a novel framework designed to overcome the limitations of 2D-based Preference-Based Reinforcement Learning (PBRL) by unifying 3D point cloud modeling and future-aware preference refinement. At its core, PRISM adopts a 3D Point Cloud-Language Model (3D-PC-LLM) to mitigate occlusion and viewpoint biases, ensuring more stable and spatially consistent preference signals. Additionally, PRISM leverages Chain-of-Thought (CoT) reasoning to incorporate long-horizon considerations, thereby preventing the short-sighted feedback often seen in static preference comparisons. In contrast to conventional PBRL techniques, this integration of 3D perception and future-oriented reasoning leads to significant gains in preference agreement rates, faster policy convergence, and robust generalization across unseen robotic environments. Our empirical results, spanning tasks such as robotic manipulation and autonomous navigation, highlight PRISM's potential for real-world applications where precise spatial understanding and reliable long-term decision-making are critical. By bridging 3D geometric awareness with CoT-driven preference modeling, PRISM establishes a comprehensive foundation for scalable, human-aligned reinforcement learning.

Yanjun Chen, Xinming Zhang, Xianghui Wang et al.

Soft Actor-Critic algorithm is widely recognized for its robust performance across a range of deep reinforcement learning tasks, where it leverages the tanh transformation to constrain actions within bounded limits. However, this transformation induces a distribution shift, distorting the original Gaussian action distribution and potentially leading the policy to select suboptimal actions, particularly in high-dimensional action spaces. In this paper, we conduct a comprehensive theoretical and empirical analysis of this distribution shift, deriving the precise probability density function (PDF) for actions following the tanh transformation to clarify the misalignment introduced between the transformed distribution's mode and the intended action output. We substantiate these theoretical insights through extensive experiments on high-dimensional tasks within the HumanoidBench benchmark. Our findings indicate that accounting for this distribution shift substantially enhances SAC's performance, resulting in notable improvements in cumulative rewards, sample efficiency, and reliability across tasks. These results underscore a critical consideration for SAC and similar algorithms: addressing transformation-induced distribution shifts is essential to optimizing policy effectiveness in high-dimensional deep reinforcement learning environments, thereby expanding the robustness and applicability of SAC in complex control tasks.

Hao Zhou, Chongyang Zhang, Yan Luo et al.

Temporal grounding aims to localize temporal boundaries within untrimmed videos by language queries, but it faces the challenge of two types of inevitable human uncertainties: query uncertainty and label uncertainty. The two uncertainties stem from human subjectivity, leading to limited generalization ability of temporal grounding. In this work, we propose a novel DeNet (Decoupling and De-bias) to embrace human uncertainty: Decoupling - We explicitly disentangle each query into a relation feature and a modified feature. The relation feature, which is mainly based on skeleton-like words (including nouns and verbs), aims to extract basic and consistent information in the presence of query uncertainty. Meanwhile, modified feature assigned with style-like words (including adjectives, adverbs, etc) represents the subjective information, and thus brings personalized predictions; De-bias - We propose a de-bias mechanism to generate diverse predictions, aim to alleviate the bias caused by single-style annotations in the presence of label uncertainty. Moreover, we put forward new multi-label metrics to diversify the performance evaluation. Extensive experiments show that our approach is more effective and robust than state-of-the-arts on Charades-STA and ActivityNet Captions datasets.