Beibin Li

Qingyun Wu, Gagan Bansal, Jieyu Zhang et al. · uw

AutoGen is an open-source framework that allows developers to build LLM applications via multiple agents that can converse with each other to accomplish tasks. AutoGen agents are customizable, conversable, and can operate in various modes that employ combinations of LLMs, human inputs, and tools. Using AutoGen, developers can also flexibly define agent interaction behaviors. Both natural language and computer code can be used to program flexible conversation patterns for different applications. AutoGen serves as a generic infrastructure to build diverse applications of various complexities and LLM capacities. Empirical studies demonstrate the effectiveness of the framework in many example applications, with domains ranging from mathematics, coding, question answering, operations research, online decision-making, entertainment, etc.

Beibin Li, Konstantina Mellou, Bo Zhang et al. · microsoft-research, uw

Supply chain operations traditionally involve a variety of complex decision making problems. Over the last few decades, supply chains greatly benefited from advances in computation, which allowed the transition from manual processing to automation and cost-effective optimization. Nonetheless, business operators still need to spend substantial efforts in explaining and interpreting the optimization outcomes to stakeholders. Motivated by the recent advances in Large Language Models (LLMs), we study how this disruptive technology can help bridge the gap between supply chain automation and human comprehension and trust thereof. We design OptiGuide -- a framework that accepts as input queries in plain text, and outputs insights about the underlying optimization outcomes. Our framework does not forgo the state-of-the-art combinatorial optimization technology, but rather leverages it to quantitatively answer what-if scenarios (e.g., how would the cost change if we used supplier B instead of supplier A for a given demand?). Importantly, our design does not require sending proprietary data over to LLMs, which can be a privacy concern in some circumstances. We demonstrate the effectiveness of our framework on a real server placement scenario within Microsoft's cloud supply chain. Along the way, we develop a general evaluation benchmark, which can be used to evaluate the accuracy of the LLM output in other scenarios.

Yun He, Wenzhe Li, Hejia Zhang et al.

Recent progress in large language models (LLMs) has led to impressive performance on a range of tasks, yet advanced instruction following (IF)-especially for complex, multi-turn, and system-prompted instructions-remains a significant challenge. Rigorous evaluation and effective training for such capabilities are hindered by the lack of high-quality, human-annotated benchmarks and reliable, interpretable reward signals. In this work, we introduce AdvancedIF (we will release this benchmark soon), a comprehensive benchmark featuring over 1,600 prompts and expert-curated rubrics that assess LLMs ability to follow complex, multi-turn, and system-level instructions. We further propose RIFL (Rubric-based Instruction-Following Learning), a novel post-training pipeline that leverages rubric generation, a finetuned rubric verifier, and reward shaping to enable effective reinforcement learning for instruction following. Extensive experiments demonstrate that RIFL substantially improves the instruction-following abilities of LLMs, achieving a 6.7% absolute gain on AdvancedIF and strong results on public benchmarks. Our ablation studies confirm the effectiveness of each component in RIFL. This work establishes rubrics as a powerful tool for both training and evaluating advanced IF in LLMs, paving the way for more capable and reliable AI systems.

Shaokun Zhang, Ming Yin, Jieyu Zhang et al.

Failure attribution in LLM multi-agent systems-identifying the agent and step responsible for task failures-provides crucial clues for systems debugging but remains underexplored and labor-intensive. In this paper, we propose and formulate a new research area: automated failure attribution for LLM multi-agent systems. To support this initiative, we introduce the Who&When dataset, comprising extensive failure logs from 127 LLM multi-agent systems with fine-grained annotations linking failures to specific agents and decisive error steps. Using the Who&When, we develop and evaluate three automated failure attribution methods, summarizing their corresponding pros and cons. The best method achieves 53.5% accuracy in identifying failure-responsible agents but only 14.2% in pinpointing failure steps, with some methods performing below random. Even SOTA reasoning models, such as OpenAI o1 and DeepSeek R1, fail to achieve practical usability. These results highlight the task's complexity and the need for further research in this area. Code and dataset are available at https://github.com/mingyin1/Agents_Failure_Attribution

Runlong Zhou, Simon S. Du, Beibin Li · tsinghua

As language models (LMs) demonstrate their capabilities in various fields, their application to tasks requiring multi-round interactions has become increasingly popular. These tasks usually have complex dynamics, so supervised fine-tuning (SFT) on a limited offline dataset does not yield good performance. However, only a few works attempted to directly train the LMs within interactive decision-making environments. We aim to create an effective approach to fine-tune LMs with online reinforcement learning (RL) in these environments. We propose Reflect-RL, a two-player system to fine-tune an LM using SFT and online RL, where a frozen reflection model (player) assists the policy model (player). To generate data for the warm-up SFT stage, we use negative example generation to enhance the error-correction ability of the reflection model. Furthermore, we designed single-prompt action enumeration and applied curriculum learning to allow the policy model to learn more efficiently. Empirically, we verify that Reflect-RL outperforms SFT and online RL without reflection. Testing results indicate GPT-2 XL 1.56B fine-tuned with Reflect-RL outperforms larger open-source LMs, such as Mistral 7B. The benchmarks, dataset, and code involved in this work are publicly available: https://github.com/zhourunlong/Reflect-RL.

Beibin Li, Nicholas Nuechterlein, Erin Barney et al.

Identifying oculomotor behaviors relevant for eye-tracking applications is a critical but often challenging task. Aiming to automatically learn and extract knowledge from existing eye-tracking data, we develop a novel method that creates rich representations of oculomotor scanpaths to facilitate the learning of downstream tasks. The proposed stimulus-agnostic Oculomotor Behavior Framework (OBF) model learns human oculomotor behaviors from unsupervised and semi-supervised tasks, including reconstruction, predictive coding, fixation identification, and contrastive learning tasks. The resultant pre-trained OBF model can be used in a variety of applications. Our pre-trained model outperforms baseline approaches and traditional scanpath methods in autism spectrum disorder and viewed-stimulus classification tasks. Ablation experiments further show our proposed method could achieve even better results with larger model sizes and more diverse eye-tracking training datasets, supporting the model's potential for future eye-tracking applications. Open source code: http://github.com/BeibinLi/OBF.

Fatemeh Ghezloo, Mehmet Saygin Seyfioglu, Rustin Soraki et al.

Diagnosing diseases through histopathology whole slide images (WSIs) is fundamental in modern pathology but is challenged by the gigapixel scale and complexity of WSIs. Trained histopathologists overcome this challenge by navigating the WSI, looking for relevant patches, taking notes, and compiling them to produce a final holistic diagnostic. Traditional AI approaches, such as multiple instance learning and transformer-based models, fail short of such a holistic, iterative, multi-scale diagnostic procedure, limiting their adoption in the real-world. We introduce PathFinder, a multi-modal, multi-agent framework that emulates the decision-making process of expert pathologists. PathFinder integrates four AI agents, the Triage Agent, Navigation Agent, Description Agent, and Diagnosis Agent, that collaboratively navigate WSIs, gather evidence, and provide comprehensive diagnoses with natural language explanations. The Triage Agent classifies the WSI as benign or risky; if risky, the Navigation and Description Agents iteratively focus on significant regions, generating importance maps and descriptive insights of sampled patches. Finally, the Diagnosis Agent synthesizes the findings to determine the patient's diagnostic classification. Our Experiments show that PathFinder outperforms state-of-the-art methods in skin melanoma diagnosis by 8% while offering inherent explainability through natural language descriptions of diagnostically relevant patches. Qualitative analysis by pathologists shows that the Description Agent's outputs are of high quality and comparable to GPT-4o. PathFinder is also the first AI-based system to surpass the average performance of pathologists in this challenging melanoma classification task by 9%, setting a new record for efficient, accurate, and interpretable AI-assisted diagnostics in pathology. Data, code and models available at https://pathfinder-dx.github.io/

Guanghao Ye, Khiem Duc Pham, Xinzhi Zhang et al.

Recent AI advancements, such as OpenAI's new models, are transforming LLMs into LRMs (Large Reasoning Models) that perform reasoning during inference, taking extra time and compute for higher-quality outputs. We aim to uncover the algorithmic framework for training LRMs. Methods like self-consistency, PRM, and AlphaZero suggest reasoning as guided search. We ask: what is the simplest, most scalable way to enable search in LLMs? We propose a post-training framework called Reinforcement Learning via Self-Play (RLSP). RLSP involves three steps: (1) supervised fine-tuning with human or synthetic demonstrations of the reasoning process, (2) using an exploration reward signal to encourage diverse and efficient reasoning behaviors, and (3) RL training with an outcome verifier to ensure correctness while preventing reward hacking. Our key innovation is to decouple exploration and correctness signals during PPO training, carefully balancing them to improve performance and efficiency. Empirical studies in the math domain show that RLSP improves reasoning. On the Llama-3.1-8B-Instruct model, RLSP can boost performance by 23% in MATH-500 test set; On AIME 2024 math problems, Qwen2.5-32B-Instruct improved by 10% due to RLSP. However, a more important finding of this work is that the models trained using RLSP, even with the simplest exploration reward that encourages the model to take more intermediate steps, showed several emergent behaviors such as backtracking, exploration of ideas, and verification. These findings demonstrate that RLSP framework might be enough to enable emergence of complex reasoning abilities in LLMs when scaled. Lastly, we propose a theory as to why RLSP search strategy is more suitable for LLMs inspired by a remarkable result that says CoT provably increases computational power of LLMs, which grows as the number of steps in CoT \cite{li2024chain,merrill2023expresssive}.

Yao Lu, Song Bian, Lequn Chen et al.

In this paper, we investigate the intersection of large generative AI models and cloud-native computing architectures. Recent large models such as ChatGPT, while revolutionary in their capabilities, face challenges like escalating costs and demand for high-end GPUs. Drawing analogies between large-model-as-a-service (LMaaS) and cloud database-as-a-service (DBaaS), we describe an AI-native computing paradigm that harnesses the power of both cloud-native technologies (e.g., multi-tenancy and serverless computing) and advanced machine learning runtime (e.g., batched LoRA inference). These joint efforts aim to optimize costs-of-goods-sold (COGS) and improve resource accessibility. The journey of merging these two domains is just at the beginning and we hope to stimulate future research and development in this area.

Beibin Li, Yi Zhang, Sébastien Bubeck et al. · microsoft-research, uw

We study the efficacy of Small Language Models (SLMs) in facilitating application usage through natural language interactions. Our focus here is on a particular internal application used in Microsoft for cloud supply chain fulfilment. Our experiments show that small models can outperform much larger ones in terms of both accuracy and running time, even when fine-tuned on small datasets. Alongside these results, we also highlight SLM-based system design considerations.

Yuyang Huang, Yuhan Liu, Haryadi S. Gunawi et al.

Continual learning has become a promising solution to refine large language models incrementally by leveraging user feedback. In particular, online continual learning - iteratively training the model with small batches of user feedback - has demonstrated notable performance improvements. However, the existing practice of separating training and serving processes forces the online trainer to recompute the intermediate results already done during serving. Such redundant computations can account for 30%-42% of total training time. In this paper, we propose Alchemist, to the best of our knowledge, the first online continual learning system that efficiently reuses serving activations to increase training throughput. Alchemist introduces two key techniques: (1) recording and storing activations and KV cache only during the prefill phase to minimize latency and memory overhead; and (2) smart activation offloading and hedging. Evaluations with inputs of varied token length sampled from ShareGPT dataset show that compared with a separate training cluster, Alchemist significantly increases training throughput by up to 1.72x, reduces up to 47% memory usage during training, and supports up to 2x more training tokens - all while maintaining negligible impact on serving latency.

Pantea Karimi, Solal Pirelli, Siva Kesava Reddy Kakarla et al.

Many problems that cloud operators solve are computationally expensive, and operators often use heuristic algorithms (that are faster and scale better than optimal) to solve them more efficiently. Heuristic analyzers enable operators to find when and by how much their heuristics underperform. However, these tools do not provide enough detail for operators to mitigate the heuristic's impact in practice: they only discover a single input instance that causes the heuristic to underperform (and not the full set), and they do not explain why. We propose XPlain, a tool that extends these analyzers and helps operators understand when and why their heuristics underperform. We present promising initial results that show such an extension is viable.

Shengyu Feng, Yun He, Shuang Ma et al.

Reinforcement learning (RL) has recently proven effective at scaling chain-of-thought (CoT) reasoning in large language models on tasks with verifiable answers. However, extending RL to more general non-verifiable tasks, typically in the format of human preference pairs, remains both challenging and underexplored. In this work, we propose Dual-Weighted Reinforcement Learning (DWRL), a new framework for preference modeling that integrates CoT reasoning with the Bradley-Terry (BT) model via a dual-weighted RL objective that preserves preference-modeling inductive bias. DWRL approximates the maximum-likelihood objective of the BT model with two complementary weights: an instance-wise misalignment weight, which emphasizes under-trained pairs misaligned with human preference, and a group-wise (self-normalized) conditional preference score, which promotes promising thoughts. In this paper, we apply DWRL to preference modeling by training generative preference models (GPMs) to first generate a thought and then predict the human preference score. Across multiple benchmarks and model scales (Llama3 and Qwen2.5), DWRL consistently outperforms both GPM baselines and scalar models, while producing coherent, interpretable thoughts. In summary, our results position DWRL as a general framework for reasoning-enhanced preference learning beyond verifiable tasks.

Beibin Li, Ezgi Mercan, Sachin Mehta et al.

In this study, we propose the Ductal Instance-Oriented Pipeline (DIOP) that contains a duct-level instance segmentation model, a tissue-level semantic segmentation model, and three-levels of features for diagnostic classification. Based on recent advancements in instance segmentation and the Mask R-CNN model, our duct-level segmenter tries to identify each ductal individual inside a microscopic image; then, it extracts tissue-level information from the identified ductal instances. Leveraging three levels of information obtained from these ductal instances and also the histopathology image, the proposed DIOP outperforms previous approaches (both feature-based and CNN-based) in all diagnostic tasks; for the four-way classification task, the DIOP achieves comparable performance to general pathologists in this unique dataset. The proposed DIOP only takes a few seconds to run in the inference time, which could be used interactively on most modern computers. More clinical explorations are needed to study the robustness and generalizability of this system in the future.

Beibin Li, Nicholas Nuechterlein, Erin Barney et al.

In genomic analysis, biomarker discovery, image recognition, and other systems involving machine learning, input variables can often be organized into different groups by their source or semantic category. Eliminating some groups of variables can expedite the process of data acquisition and avoid over-fitting. Researchers have used the group lasso to ensure group sparsity in linear models and have extended it to create compact neural networks in meta-learning. Different from previous studies, we use multi-layer non-linear neural networks to find sparse groups for input variables. We propose a new loss function to regularize parameters for grouped input variables, design a new optimization algorithm for this loss function, and test these methods in three real-world settings. We achieve group sparsity for three datasets, maintaining satisfying results while excluding one nucleotide position from an RNA splicing experiment, excluding 89.9% of stimuli from an eye-tracking experiment, and excluding 60% of image rows from an experiment on the MNIST dataset.

Beibin Li, Sachin Mehta, Deepali Aneja et al.

In this paper, we introduce an end-to-end machine learning-based system for classifying autism spectrum disorder (ASD) using facial attributes such as expressions, action units, arousal, and valence. Our system classifies ASD using representations of different facial attributes from convolutional neural networks, which are trained on images in the wild. Our experimental results show that different facial attributes used in our system are statistically significant and improve sensitivity, specificity, and F1 score of ASD classification by a large margin. In particular, the addition of different facial attributes improves the performance of ASD classification by about 7% which achieves a F1 score of 76%.