SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning

Recent advances in inference-time compute have significantly improved performance on complex tasks by generating long chains of thought (CoTs) using Large Reasoning Models (LRMs). However, this improved accuracy comes at the cost of high inference latency due to the length of generated reasoning sequences and the autoregressive nature of decoding. Our key insight in tackling these overheads is that LRM inference, and the reasoning that it embeds, is highly tolerant of approximations: complex tasks are typically broken down into simpler steps, each of which brings utility based on the semantic insight it provides for downstream steps rather than the exact tokens it generates. Accordingly, we introduce SpecReason, a system that automatically accelerates LRM inference by using a lightweight model to (speculatively) carry out simpler intermediate reasoning steps and reserving the costly base model only to assess (and potentially correct) the speculated outputs. Importantly, SpecReason's focus on exploiting the semantic flexibility of thinking tokens in preserving final-answer accuracy is complementary to prior speculation techniques, most notably speculative decoding, which demands token-level equivalence at each step. Across a variety of reasoning benchmarks, SpecReason achieves 1.5-2.5$\times$ speedup over vanilla LRM inference while improving accuracy by 1.0-9.9\%. Compared to speculative decoding without SpecReason, their combination yields an additional 19.4-44.2\% latency reduction. We open-source SpecReason at https://github.com/ruipeterpan/specreason.

PFDial: A Structured Dialogue Instruction Fine-tuning Method Based on UML Flowcharts

Process-driven dialogue systems, which operate under strict predefined process constraints, are essential in customer service and equipment maintenance scenarios. Although Large Language Models (LLMs) have shown remarkable progress in dialogue and reasoning, they still struggle to solve these strictly constrained dialogue tasks. To address this challenge, we construct Process Flow Dialogue (PFDial) dataset, which contains 12,705 high-quality Chinese dialogue instructions derived from 440 flowcharts containing 5,055 process nodes. Based on PlantUML specification, each UML flowchart is converted into atomic dialogue units i.e., structured five-tuples. Experimental results demonstrate that a 7B model trained with merely 800 samples, and a 0.5B model trained on total data both can surpass 90% accuracy. Additionally, the 8B model can surpass GPT-4o up to 43.88% with an average of 11.00%. We further evaluate models' performance on challenging backward transitions in process flows and conduct an in-depth analysis of various dataset formats to reveal their impact on model performance in handling decision and sequential branches. The data is released in https://github.com/KongLongGeFDU/PFDial.

Real-time Spatial-temporal Traversability Assessment via Feature-based Sparse Gaussian Process

Terrain analysis is critical for the practical application of ground mobile robots in real-world tasks, especially in outdoor unstructured environments. In this paper, we propose a novel spatial-temporal traversability assessment method, which aims to enable autonomous robots to effectively navigate through complex terrains. Our approach utilizes sparse Gaussian processes (SGP) to extract geometric features (curvature, gradient, elevation, etc.) directly from point cloud scans. These features are then used to construct a high-resolution local traversability map. Then, we design a spatial-temporal Bayesian Gaussian kernel (BGK) inference method to dynamically evaluate traversability scores, integrating historical and real-time data while considering factors such as slope, flatness, gradient, and uncertainty metrics. GPU acceleration is applied in the feature extraction step, and the system achieves real-time performance. Extensive simulation experiments across diverse terrain scenarios demonstrate that our method outperforms SOTA approaches in both accuracy and computational efficiency. Additionally, we develop an autonomous navigation framework integrated with the traversability map and validate it with a differential driven vehicle in complex outdoor environments. Our code will be open-source for further research and development by the community, https://github.com/ZJU-FAST-Lab/FSGP_BGK.

SEB-Naver: A SE(2)-based Local Navigation Framework for Car-like Robots on Uneven Terrain

Autonomous navigation of car-like robots on uneven terrain poses unique challenges compared to flat terrain, particularly in traversability assessment and terrain-associated kinematic modelling for motion planning. This paper introduces SEB-Naver, a novel SE(2)-based local navigation framework designed to overcome these challenges. First, we propose an efficient traversability assessment method for SE(2) grids, leveraging GPU parallel computing to enable real-time updates and maintenance of local maps. Second, inspired by differential flatness, we present an optimization-based trajectory planning method that integrates terrain-associated kinematic models, significantly improving both planning efficiency and trajectory quality. Finally, we unify these components into SEB-Naver, achieving real-time terrain assessment and trajectory optimization. Extensive simulations and real-world experiments demonstrate the effectiveness and efficiency of our approach. The code is at https://github.com/ZJU-FAST-Lab/seb_naver.

EliteKV: Scalable KV Cache Compression via RoPE Frequency Selection and Joint Low-Rank Projection

Rotary Position Embedding (RoPE) enables each attention head to capture multi-frequency information along the sequence dimension and is widely applied in foundation models. However, the nonlinearity introduced by RoPE complicates optimization of the key state in the Key-Value (KV) cache for RoPE-based attention. Existing KV cache compression methods typically store key state before rotation and apply the transformation during decoding, introducing additional computational overhead. This paper introduces EliteKV, a flexible modification framework for RoPE-based models supporting variable KV cache compression ratios. EliteKV first identifies the intrinsic frequency preference of each head using RoPElite, selectively restoring linearity to certain dimensions of key within attention computation. Building on this, joint low-rank compression of key and value enables partial cache sharing. Experimental results show that with minimal uptraining on only $0.6\%$ of the original training data, RoPE-based models achieve a $75\%$ reduction in KV cache size while preserving performance within a negligible margin. Furthermore, EliteKV consistently performs well across models of different scales within the same family.

Extensive Exploration in Complex Traffic Scenarios using Hierarchical Reinforcement Learning

Developing an automated driving system capable of navigating complex traffic environments remains a formidable challenge. Unlike rule-based or supervised learning-based methods, Deep Reinforcement Learning (DRL) based controllers eliminate the need for domain-specific knowledge and datasets, thus providing adaptability to various scenarios. Nonetheless, a common limitation of existing studies on DRL-based controllers is their focus on driving scenarios with simple traffic patterns, which hinders their capability to effectively handle complex driving environments with delayed, long-term rewards, thus compromising the generalizability of their findings. In response to these limitations, our research introduces a pioneering hierarchical framework that efficiently decomposes intricate decision-making problems into manageable and interpretable subtasks. We adopt a two step training process that trains the high-level controller and low-level controller separately. The high-level controller exhibits an enhanced exploration potential with long-term delayed rewards, and the low-level controller provides longitudinal and lateral control ability using short-term instantaneous rewards. Through simulation experiments, we demonstrate the superiority of our hierarchical controller in managing complex highway driving situations.

AdaServe: SLO-Customized LLM Serving with Fine-Grained Speculative Decoding

This paper introduces AdaServe, the first LLM serving system to support SLO customization through fine-grained speculative decoding. AdaServe leverages the logits of a draft model to predict the speculative accuracy of tokens and employs a theoretically optimal algorithm to construct token trees for verification. To accommodate diverse SLO requirements without compromising throughput, AdaServe employs a speculation-and-selection scheme that first constructs candidate token trees for each request and then dynamically selects tokens to meet individual SLO constraints while optimizing throughput. Comprehensive evaluations demonstrate that AdaServe achieves up to 73% higher SLO attainment and 74% higher goodput compared to state-of-the-art systems. These results underscore AdaServe's potential to enhance the efficiency and adaptability of LLM deployments across varied application scenarios.

Communication Bounds for the Distributed Experts Problem

In this work, we study the experts problem in the distributed setting where an expert's cost needs to be aggregated across multiple servers. Our study considers various communication models such as the message-passing model and the broadcast model, along with multiple aggregation functions, such as summing and taking the $\ell_p$ norm of an expert's cost across servers. We propose the first communication-efficient protocols that achieve near-optimal regret in these settings, even against a strong adversary who can choose the inputs adaptively. Additionally, we give a conditional lower bound showing that the communication of our protocols is nearly optimal. Finally, we implement our protocols and demonstrate empirical savings on the HPO-B benchmarks.

LLMs can realize combinatorial creativity: generating creative ideas via LLMs for scientific research

Scientific idea generation has been extensively studied in creativity theory and computational creativity research, providing valuable frameworks for understanding and implementing creative processes. However, recent work using Large Language Models (LLMs) for research idea generation often overlooks these theoretical foundations. We present a framework that explicitly implements combinatorial creativity theory using LLMs, featuring a generalization-level retrieval system for cross-domain knowledge discovery and a structured combinatorial process for idea generation. The retrieval system maps concepts across different abstraction levels to enable meaningful connections between disparate domains, while the combinatorial process systematically analyzes and recombines components to generate novel solutions. Experiments on the OAG-Bench dataset demonstrate our framework's effectiveness, consistently outperforming baseline approaches in generating ideas that align with real research developments (improving similarity scores by 7\%-10\% across multiple metrics). Our results provide strong evidence that LLMs can effectively realize combinatorial creativity when guided by appropriate theoretical frameworks, contributing both to practical advancement of AI-assisted research and theoretical understanding of machine creativity.

A Claim Decomposition Benchmark for Long-form Answer Verification

The advancement of LLMs has significantly boosted the performance of complex long-form question answering tasks. However, one prominent issue of LLMs is the generated "hallucination" responses that are not factual. Consequently, attribution for each claim in responses becomes a common solution to improve the factuality and verifiability. Existing researches mainly focus on how to provide accurate citations for the response, which largely overlook the importance of identifying the claims or statements for each response. To bridge this gap, we introduce a new claim decomposition benchmark, which requires building system that can identify atomic and checkworthy claims for LLM responses. Specifically, we present the Chinese Atomic Claim Decomposition Dataset (CACDD), which builds on the WebCPM dataset with additional expert annotations to ensure high data quality. The CACDD encompasses a collection of 500 human-annotated question-answer pairs, including a total of 4956 atomic claims. We further propose a new pipeline for human annotation and describe the challenges of this task. In addition, we provide experiment results on zero-shot, few-shot and fine-tuned LLMs as baselines. The results show that the claim decomposition is highly challenging and requires further explorations. All code and data are publicly available at \url{https://github.com/FBzzh/CACDD}.