Jianuo Huang

Jianuo Huang, Yaojie Zhang, Qituan Zhang et al.

Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequential overhead, while parallel drafters reduce drafting cost but weaken intra-block dependency modeling. In this paper, we propose Domino, a speculative decoding framework that decouples causal dependency modeling from expensive autoregressive draft execution. Domino first uses a parallel draft backbone to produce preliminary draft distributions for the entire block, and then applies a lightweight Domino head to refine them with prefix-dependent causal information. To stabilize teacher-forced causal encoding, we further introduce a base-anchored training curriculum that first strengthens the parallel backbone and then gradually shifts optimization toward the causally corrected final distribution. Experiments on Qwen3 models show that Domino achieves up to \(5.49\times\) end-to-end speedup under the Transformers backend and up to \(5.8\times\) throughput speedup under SGLang serving.

Yaojie Zhang, Jianuo Huang, Junlong Ke et al.

Speculative decoding accelerates memory-bound LLM inference without quality degradation by using a fast drafter to propose multiple candidate tokens and the target model to verify them in parallel. However, conventional sequential speculative decoding suffers from mutual waiting between drafting and verification, and repeated exchange of intermediate states further increases memory access overhead. Parallel speculative decoding addresses this limitation by performing drafting and verification within a single target forward pass, allowing future drafts to be prepared while current candidates are being verified. Although effective at small batch sizes, existing parallel speculative decoding methods either require costly continual pretraining with quality degradation or suffer from low acceptance rates. More importantly, this paradigm inherently suffers from uncertainty in both the bonus token and the accepted length, leading to draft verification mismatch and causing throughput gains to collapse at large batch sizes. To address these limitations, we introduce FlexDraft, a lossless speculative decoding framework that flexibly adapts to varying batch sizes through three key designs. (1) Attention Tuning enables block diffusion drafting by tuning only the attention projectors of the final few layers on mask tokens, while keeping the autoregressive path frozen to preserve the target distribution and produce high quality drafts with minimal trainable parameters. (2) Bonus-guided Calibration uses a lightweight MLP conditioned on the resolved bonus token to calibrate draft logits, mitigating draft verification mismatch caused by bonus token uncertainty. (3) Flex Decoding dynamically switches between parallel draft and verify at small batch sizes and sequential draft then verify at large batch sizes, and adjusts verification length based on draft confidence to eliminate redundant computation.

Jianuo Huang, Yaojie Zhang, Yicun Yang et al. · cmu

Diffusion large language models (dLLMs) present a promising alternative to dominant autoregressive models (ARMs) by the ability of parallel decoding at the expense of substantial computation and memory costs. Specifically, the cache mechanism for bidirectional attention in dLLMs demands large memory footprint, restricting their ability to handle long contexts under resource-limited settings. Existing cache eviction strategies are designed for ARMs and ignore the unique characteristics of dLLMs, thus leading to unsatisfactory performance. To address these challenges, we introduce MaskKV, a training-free cache eviction framework tailored to dLLMs, focusing on the effect of mask tokens in dLLMs. MaskKV is built on two key innovations: (1) a mask-query guided scoring mechanism that leverages attention weights to identify and evict less critical prompt tokens for each head; (2) an adaptive cache budgeting strategy that improves efficiency by reducing allocation in intermediate layers and concentrating resources on prompt-preferring heads. On LLaDA with MaskKV, compressing the KV cache to only 256 pairs (less than 5% of tokens) retains 94% of the full-cache performance on LongBench and achieves up to 31x acceleration at 32k prompt length. The code is publicly available at: https://github.com/jianuo-huang/MaskKV

Detian Chu, Linyuan Bai, Jianuo Huang et al.

With the advancement of autonomous driving, ensuring safety during motion planning and navigation is becoming more and more important. However, most end-to-end planning methods suffer from a lack of safety. This research addresses the safety issue in the control optimization problem of autonomous driving, formulated as Constrained Markov Decision Processes (CMDPs). We propose a novel, model-based approach for policy optimization, utilizing a conditional Value-at-Risk based Soft Actor Critic to manage constraints in complex, high-dimensional state spaces effectively. Our method introduces a worst-case actor to guide safe exploration, ensuring rigorous adherence to safety requirements even in unpredictable scenarios. The policy optimization employs the Augmented Lagrangian method and leverages latent diffusion models to predict and simulate future trajectories. This dual approach not only aids in navigating environments safely but also refines the policy's performance by integrating distribution modeling to account for environmental uncertainties. Empirical evaluations conducted in both simulated and real environment demonstrate that our approach outperforms existing methods in terms of safety, efficiency, and decision-making capabilities.

Zhangjun Zhou, Yiping Li, Chunlin Zhong et al.

While the human visual system employs distinct mechanisms to perceive salient and camouflaged objects, existing models struggle to disentangle these tasks. Specifically, salient object detection (SOD) models frequently misclassify camouflaged objects as salient, while camouflaged object detection (COD) models conversely misinterpret salient objects as camouflaged. We hypothesize that this can be attributed to two factors: (i) the specific annotation paradigm of current SOD and COD datasets, and (ii) the lack of explicit attribute relationship modeling in current models. Prevalent SOD/COD datasets enforce a mutual exclusivity constraint, assuming scenes contain either salient or camouflaged objects, which poorly aligns with the real world. Furthermore, current SOD/COD methods are primarily designed for these highly constrained datasets and lack explicit modeling of the relationship between salient and camouflaged objects. In this paper, to promote the development of unconstrained salient and camouflaged object detection, we construct a large-scale dataset, USC12K, which features comprehensive labels and four different scenes that cover all possible logical existence scenarios of both salient and camouflaged objects. To explicitly model the relationship between salient and camouflaged objects, we propose a model called USCNet, which introduces two distinct prompt query mechanisms for modeling inter-sample and intra-sample attribute relationships. Additionally, to assess the model's ability to distinguish between salient and camouflaged objects, we design an evaluation metric called CSCS. The proposed method achieves state-of-the-art performance across all scenes in various metrics. The code and dataset will be available at https://github.com/ssecv/USCNet.

Jinwei Su, Yinghui Xia, Yiqun Duan et al.

Large language models (LLMs) have demonstrated strong potential and impressive performance in automating the generation and optimization of workflows. However, existing approaches are marked by limited reasoning capabilities, high computational demands, and significant resource requirements. To address these issues, we propose DebFlow, a framework that employs a debate mechanism to optimize workflows and integrates reflexion to improve based on previous experiences. We evaluated our method across six benchmark datasets, including HotpotQA, MATH, and ALFWorld. Our approach achieved a 3\% average performance improvement over the latest baselines, demonstrating its effectiveness in diverse problem domains. In particular, during training, our framework reduces resource consumption by 37\% compared to the state-of-the-art baselines. Additionally, we performed ablation studies. Removing the Debate component resulted in a 4\% performance drop across two benchmark datasets, significantly greater than the 2\% drop observed when the Reflection component was removed. These findings strongly demonstrate the critical role of Debate in enhancing framework performance, while also highlighting the auxiliary contribution of reflexion to overall optimization.

Jianuo Huang

In recent advancements in Multi-agent Reinforcement Learning (MARL), its application has extended to various safety-critical scenarios. However, most methods focus on online learning, which presents substantial risks when deployed in real-world settings. Addressing this challenge, we introduce an innovative framework integrating diffusion models within the MARL paradigm. This approach notably enhances the safety of actions taken by multiple agents through risk mitigation while modeling coordinated action. Our framework is grounded in the Centralized Training with Decentralized Execution (CTDE) architecture, augmented by a Diffusion Model for prediction trajectory generation. Additionally, we incorporate a specialized algorithm to further ensure operational safety. We evaluate our model against baselines on the DSRL benchmark. Experiment results demonstrate that our model not only adheres to stringent safety constraints but also achieves superior performance compared to existing methodologies. This underscores the potential of our approach in advancing the safety and efficacy of MARL in real-world applications.