MedAgentGym: Training LLM Agents for Code-Based Medical Reasoning at Scale

We introduce MedAgentGYM, the first publicly available training environment designed to enhance coding-based medical reasoning capabilities in large language model (LLM) agents. MedAgentGYM comprises 72,413 task instances across 129 categories derived from authentic real-world biomedical scenarios. Tasks are encapsulated within executable coding environments, each featuring detailed task descriptions, interactive feedback mechanisms, verifiable ground-truth annotations, and scalable training trajectory generation. Extensive benchmarking of over 30 LLMs reveals a notable performance disparity between commercial API-based models and open-source counterparts. Leveraging MedAgentGYM, Med-Copilot-7B achieves substantial performance gains through supervised fine-tuning (+36.44%) and continued reinforcement learning (+42.47%), emerging as an affordable and privacy-preserving alternative competitive with gpt-4o. By offering both a comprehensive benchmark and accessible, expandable training resources within unified execution environments, MedAgentGYM delivers an integrated platform to develop LLM-based coding assistants for advanced biomedical research and practice.

SpecRouter: Adaptive Routing for Multi-Level Speculative Decoding in Large Language Models

Large Language Models (LLMs) present a critical trade-off between inference quality and computational cost: larger models offer superior capabilities but incur significant latency, while smaller models are faster but less powerful. Existing serving strategies often employ fixed model scales or static two-stage speculative decoding, failing to dynamically adapt to the varying complexities of user requests or fluctuations in system performance. This paper introduces \systemname{}, a novel framework that reimagines LLM inference as an adaptive routing problem solved through multi-level speculative decoding. \systemname{} dynamically constructs and optimizes inference "paths" (chains of models) based on real-time feedback, addressing the limitations of static approaches. Our contributions are threefold: (1) An \textbf{adaptive model chain scheduling} mechanism that leverages performance profiling (execution times) and predictive similarity metrics (derived from token distribution divergence) to continuously select the optimal sequence of draft and verifier models, minimizing predicted latency per generated token. (2) A \textbf{multi-level collaborative verification} framework where intermediate models within the selected chain can validate speculative tokens, reducing the verification burden on the final, most powerful target model. (3) A \textbf{synchronized state management} system providing efficient, consistent KV cache handling across heterogeneous models in the chain, including precise, low-overhead rollbacks tailored for asynchronous batch processing inherent in multi-level speculation. Preliminary experiments demonstrate the validity of our method.

BFS-Prover: Scalable Best-First Tree Search for LLM-based Automatic Theorem Proving

Recent advancements in large language models (LLMs) have spurred growing interest in automatic theorem proving using Lean4, where effective tree search methods are crucial for navigating proof search spaces. While the existing approaches primarily rely on value functions and Monte Carlo Tree Search (MCTS), the potential of simpler methods like Best-First Search (BFS) remains underexplored. This paper investigates whether BFS can achieve competitive performance in large-scale theorem proving tasks. We present \texttt{BFS-Prover}, a scalable expert iteration framework, featuring three key innovations. First, we implement strategic data filtering at each expert iteration round, excluding problems solvable via beam search node expansion to focus on harder cases. Second, we improve the sample efficiency of BFS through Direct Preference Optimization (DPO) applied to state-tactic pairs automatically annotated with compiler error feedback, refining the LLM's policy to prioritize productive expansions. Third, we employ length normalization in BFS to encourage exploration of deeper proof paths. \texttt{BFS-Prover} achieves a score of $71.31$ on the MiniF2F test set and therefore challenges the perceived necessity of complex tree search methods, demonstrating that BFS can achieve competitive performance when properly scaled.

BLOS-BEV: Navigation Map Enhanced Lane Segmentation Network, Beyond Line of Sight

Bird's-eye-view (BEV) representation is crucial for the perception function in autonomous driving tasks. It is difficult to balance the accuracy, efficiency and range of BEV representation. The existing works are restricted to a limited perception range within 50 meters. Extending the BEV representation range can greatly benefit downstream tasks such as topology reasoning, scene understanding, and planning by offering more comprehensive information and reaction time. The Standard-Definition (SD) navigation maps can provide a lightweight representation of road structure topology, characterized by ease of acquisition and low maintenance costs. An intuitive idea is to combine the close-range visual information from onboard cameras with the beyond line-of-sight (BLOS) environmental priors from SD maps to realize expanded perceptual capabilities. In this paper, we propose BLOS-BEV, a novel BEV segmentation model that incorporates SD maps for accurate beyond line-of-sight perception, up to 200m. Our approach is applicable to common BEV architectures and can achieve excellent results by incorporating information derived from SD maps. We explore various feature fusion schemes to effectively integrate the visual BEV representations and semantic features from the SD map, aiming to leverage the complementary information from both sources optimally. Extensive experiments demonstrate that our approach achieves state-of-the-art performance in BEV segmentation on nuScenes and Argoverse benchmark. Through multi-modal inputs, BEV segmentation is significantly enhanced at close ranges below 50m, while also demonstrating superior performance in long-range scenarios, surpassing other methods by over 20% mIoU at distances ranging from 50-200m.

MapLocNet: Coarse-to-Fine Feature Registration for Visual Re-Localization in Navigation Maps

Robust localization is the cornerstone of autonomous driving, especially in challenging urban environments where GPS signals suffer from multipath errors. Traditional localization approaches rely on high-definition (HD) maps, which consist of precisely annotated landmarks. However, building HD map is expensive and challenging to scale up. Given these limitations, leveraging navigation maps has emerged as a promising low-cost alternative for localization. Current approaches based on navigation maps can achieve highly accurate localization, but their complex matching strategies lead to unacceptable inference latency that fails to meet the real-time demands. To address these limitations, we propose a novel transformer-based neural re-localization method. Inspired by image registration, our approach performs a coarse-to-fine neural feature registration between navigation map and visual bird's-eye view features. Our method significantly outperforms the current state-of-the-art OrienterNet on both the nuScenes and Argoverse datasets, which is nearly 10%/20% localization accuracy and 30/16 FPS improvement on single-view and surround-view input settings, separately. We highlight that our research presents an HD-map-free localization method for autonomous driving, offering cost-effective, reliable, and scalable performance in challenging driving environments.

MedAdapter: Efficient Test-Time Adaptation of Large Language Models towards Medical Reasoning

Despite their improved capabilities in generation and reasoning, adapting large language models (LLMs) to the biomedical domain remains challenging due to their immense size and corporate privacy. In this work, we propose MedAdapter, a unified post-hoc adapter for test-time adaptation of LLMs towards biomedical applications. Instead of fine-tuning the entire LLM, MedAdapter effectively adapts the original model by fine-tuning only a small BERT-sized adapter to rank candidate solutions generated by LLMs. Experiments demonstrate that MedAdapter effectively adapts both white-box and black-box LLMs in biomedical reasoning, achieving average performance improvements of 25.48% and 11.31%, respectively, without requiring extensive computational resources or sharing data with third parties. MedAdapter also yields superior performance when combined with train-time adaptation, highlighting a flexible and complementary solution to existing adaptation methods. Faced with the challenges of balancing model performance, computational resources, and data privacy, MedAdapter provides an efficient, privacy-preserving, cost-effective, and transparent solution for adapting LLMs to the biomedical domain.

Balancing Enhancement, Harmlessness, and General Capabilities: Enhancing Conversational LLMs with Direct RLHF

In recent advancements in Conversational Large Language Models (LLMs), a concerning trend has emerged, showing that many new base LLMs experience a knowledge reduction in their foundational capabilities following Supervised Fine-Tuning (SFT). This process often leads to issues such as forgetting or a decrease in the base model's abilities. Moreover, fine-tuned models struggle to align with user preferences, inadvertently increasing the generation of toxic outputs when specifically prompted. To overcome these challenges, we adopted an innovative approach by completely bypassing SFT and directly implementing Harmless Reinforcement Learning from Human Feedback (RLHF). Our method not only preserves the base model's general capabilities but also significantly enhances its conversational abilities, while notably reducing the generation of toxic outputs. Our approach holds significant implications for fields that demand a nuanced understanding and generation of responses, such as customer service. We applied this methodology to Mistral, the most popular base model, thereby creating Mistral-Plus. Our validation across 11 general tasks demonstrates that Mistral-Plus outperforms similarly sized open-source base models and their corresponding instruct versions. Importantly, the conversational abilities of Mistral-Plus were significantly improved, indicating a substantial advancement over traditional SFT models in both safety and user preference alignment.

EHRAgent: Code Empowers Large Language Models for Complex Tabular Reasoning on Electronic Health Records

Large language models (LLMs) have demonstrated exceptional capabilities in planning and tool utilization as autonomous agents, but few have been developed for medical problem-solving. We propose EHRAgent1, an LLM agent empowered with a code interface, to autonomously generate and execute code for complex clinical tasks within electronic health records (EHRs). First, we formulate an EHR question-answering task into a tool-use planning process, efficiently decomposing a complicated task into a sequence of manageable actions. By integrating interactive coding and execution feedback, EHRAgent learns from error messages and improves the originally generated code through iterations. Furthermore, we enhance the LLM agent by incorporating long-term memory, which allows EHRAgent to effectively select and build upon the most relevant successful cases from past experiences. Experiments on two real-world EHR datasets show that EHRAgent outperforms the strongest LLM agent baseline by 36.48% and 12.41%, respectively. EHRAgent leverages the emerging few-shot learning capabilities of LLMs, enabling autonomous code generation and execution to tackle complex clinical tasks with minimal demonstrations.

6-DoF Robotic Grasping with Transformer

Robotic grasping aims to detect graspable points and their corresponding gripper configurations in a particular scene, and is fundamental for robot manipulation. Existing research works have demonstrated the potential of using a transformer model for robotic grasping, which can efficiently learn both global and local features. However, such methods are still limited in grasp detection on a 2D plane. In this paper, we extend a transformer model for 6-Degree-of-Freedom (6-DoF) robotic grasping, which makes it more flexible and suitable for tasks that concern safety. The key designs of our method are a serialization module that turns a 3D voxelized space into a sequence of feature tokens that a transformer model can consume and skip-connections that merge multiscale features effectively. In particular, our method takes a Truncated Signed Distance Function (TSDF) as input. After serializing the TSDF, a transformer model is utilized to encode the sequence, which can obtain a set of aggregated hidden feature vectors through multi-head attention. We then decode the hidden features to obtain per-voxel feature vectors through deconvolution and skip-connections. Voxel feature vectors are then used to regress parameters for executing grasping actions. On a recently proposed pile and packed grasping dataset, we showcase that our transformer-based method can surpass existing methods by about 5% in terms of success rates and declutter rates. We further evaluate the running time and generalization ability to demonstrate the superiority of the proposed method.

Public Health Informatics: Proposing Causal Sequence of Death Using Neural Machine Translation

Each year there are nearly 57 million deaths around the world, with over 2.7 million in the United States. Timely, accurate and complete death reporting is critical in public health, as institutions and government agencies rely on death reports to analyze vital statistics and to formulate responses to communicable diseases. Inaccurate death reporting may result in potential misdirection of public health policies. Determining the causes of death is, nevertheless, challenging even for experienced physicians. To facilitate physicians in accurately reporting causes of death, we present an advanced AI approach to determine a chronically ordered sequence of clinical conditions that lead to death, based on decedent's last hospital admission discharge record. The sequence of clinical codes on the death report is named as causal chain of death, coded in the tenth revision of International Statistical Classification of Diseases (ICD-10); the priority-ordered clinical conditions on the discharge record are coded in ICD-9. We identify three challenges in proposing the causal chain of death: two versions of coding system in clinical codes, medical domain knowledge conflict, and data interoperability. To overcome the first challenge in this sequence-to-sequence problem, we apply neural machine translation models to generate target sequence. We evaluate the quality of generated sequences with the BLEU (BiLingual Evaluation Understudy) score and achieve 16.44 out of 100. To address the second challenge, we incorporate expert-verified medical domain knowledge as constraint in generating output sequence to exclude infeasible causal chains. Lastly, we demonstrate the usability of our work in a Fast Healthcare Interoperability Resources (FHIR) interface to address the third challenge.