arXiv subjects

Yu-Yang Qian

Publications and source records attributed to Yu-Yang Qian.

7 recordsLinked to original sources

arXiv2026

AdaFlash: Adaptive Speculative Decoding via On-Policy Distilled Diffusion Drafters

Speculative decoding, in which a lightweight draft model first generates a draft sequence that is then verified in parallel by the target model, has become a prevalent paradigm for accelerating large language model inference. Recent work such as DFlash further boosts drafting efficiency by leveraging diffusion drafters, whose parallel denoising mechanism enables draft generation in a single forward pass. In this work, we uncover a central pitfall of diffusion drafters: bidirectional attention is a double-edged sword. On one hand, it endows the model with parallel generation and global contextual modeling capabilities; on the other hand, this inherent global dependency introduces high variance at both the domain-level and the token-level: acceptance rates fluctuate substantially across different domains, and draft token quality also varies heterogeneously at different token positions. To tackle this issue, we propose AdaFlash framework, comprising two components: (i) an on-policy distillation (OPD) algorithm with reverse-KL divergence tailored for diffusion drafters, bringing stable convergence and effectively reducing domain-level variance; and (ii) an adaptive length head that dynamically adjusts the candidate sequence length on the fly, substantially lowering the verification cost of the target model and handling token-level variance. Experiments demonstrate that AdaFlash consistently improves speedup rate during deployment, with especially significant gains in high-concurrency scenarios, achieving up to approximately 66% higher throughput than previous state-of-the-art methods.

cs.LG↗

arXiv2026

JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting

Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult to break because prior head-based SD methods face a causality-efficiency dilemma. Autoregressive drafters produce path-conditioned candidates that are effective for tree speculative decoding with higher acceptance length, but their drafting cost grows with tree depth. Bidirectional block-diffusion drafters generate all positions in one pass, but their branch-agnostic marginals can form individually plausible yet mutually inconsistent trees, wasting budget and reducing acceptance. We propose JetSpec, a head-based SD framework that combines one-forward drafting efficiency with branch-wise causal conditioning. JetSpec trains a causal parallel draft head over fused hidden states from the frozen target model, producing candidate trees whose scores align with the target model's autoregressive factorization. This enables JetSpec to convert larger draft budgets into longer accepted prefixes and higher end-to-end speedup. Across math, coding, and chat benchmarks on dense and MoE Qwen3 models, JetSpec consistently outperforms bidirectional-head and tree-based SD baselines. On H100 GPUs, JetSpec achieves up to 9.64x speedup on MATH-500 and 4.58x on open-ended conversational workloads, with further latency gains demonstrated through vLLM integration under realistic serving loads. Our code and models are available at https://github.com/hao-ai-lab/JetSpec.

cs.CL↗

arXiv2026

Robust Length Prediction: A Perspective from Heavy-Tailed Prompt-Conditioned Distributions

Output-length prediction is important for efficient LLM serving, as it directly affects batching, memory reservation, and scheduling. For prompt-only length prediction, most existing methods use a one-shot sampled length as the label, implicitly treating each prompt as if it had one true target length. We show that this is unreliable: even under a fixed model and decoding setup, the same prompt induces a \emph{prompt-conditioned output length distribution}, not a deterministic scalar, and this distribution is consistent with \emph{heavy-tailed} behavior. Motivated by this, we cast length prediction as robust estimation from heavy-tailed prompt-conditioned length distributions. We propose prompt-conditioned length distribution (ProD) methods, which construct training targets from multiple independent generations of the same prompt. Two variants are developed to reuse the served LLM's hidden states: \mbox{ProD-M}, which uses a median-based target for robust point prediction, and ProD-D, which uses a distributional target that preserves prompt-conditioned uncertainty. We provide theoretical justifications by analyzing the estimation error under a surrogate model. Experiments across diverse scenarios show consistent gains in prediction quality.

cs.LG↗

arXiv2026

When Drafts Evolve: Speculative Decoding Meets Online Learning

Speculative decoding has emerged as a widely adopted paradigm for accelerating large language model inference, where a lightweight draft model rapidly generates candidate tokens that are then verified in parallel by a larger target model. However, due to limited model capacity, drafts often struggle to approximate the target distribution, resulting in shorter acceptance lengths and diminished speedup. A key yet under-explored observation is that speculative decoding inherently provides verification feedback that quantifies the deviation between the draft and target models at no additional cost. This process naturally forms an iterative "draft commits-feedback provides-draft adapts" evolving loop, which precisely matches the online learning paradigm. Motivated by this connection, we propose OnlineSPEC, a unified framework that systematically leverages interactive feedback to continuously evolve draft models. Grounded in dynamic regret minimization, we establish a formal link between online learning performance and speculative system's acceleration rate, and develop novel algorithms via modern online learning techniques, including optimistic online learning that adaptively reuses historical gradients as predictive update hints, and online ensemble learning that dynamically maintains multiple draft models. Our algorithms are equipped with theoretical justifications and improved acceleration rates, achieving up to 24% speedup over seven benchmarks and five foundation models.

cs.LG↗

arXiv2026

d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory Distillation

Diffusion large language models (dLLMs) offer capabilities beyond those of autoregressive (AR) LLMs, such as parallel decoding and random-order generation. However, realizing these benefits in practice is non-trivial, as dLLMs inherently face an accuracy-parallelism trade-off. Despite increasing interest, existing methods typically focus on only one-side of the coin, targeting either efficiency or accuracy. To address this limitation, we propose d3LLM (Pseudo-Distilled Diffusion Large Language Model), striking a balance between accuracy and parallelism: (i) during training, we introduce pseudo-trajectory distillation to teach the model which tokens can be decoded confidently at early steps, thereby improving parallelism; (ii) during inference, we employ entropy-based multi-block decoding with a KV-cache refresh mechanism to achieve high parallelism while maintaining accuracy. To better evaluate dLLMs, we also introduce AUP (Accuracy Under Parallelism), a new metric that jointly measures accuracy and parallelism. Experiments demonstrate that our d3LLM achieves up to 10$\times$ speedup over vanilla LLaDA/Dream, and 5$\times$ speedup over AR models without much accuracy drop. Our code is available at https://github.com/hao-ai-lab/d3LLM.

cs.LG↗

arXiv2025

TreeLoRA: Efficient Continual Learning via Layer-Wise LoRAs Guided by a Hierarchical Gradient-Similarity Tree

Many real-world applications collect data in a streaming environment, where learning tasks are encountered sequentially. This necessitates continual learning (CL) to update models online, enabling adaptation to new tasks while preserving past knowledge to prevent catastrophic forgetting. Nowadays, with the flourish of large pre-trained models (LPMs), efficiency has become increasingly critical for CL, due to their substantial computational demands and growing parameter sizes. In this paper, we introduce TreeLoRA (K-D Tree of Low-Rank Adapters), a novel approach that constructs layer-wise adapters by leveraging hierarchical gradient similarity to enable efficient CL, particularly for LPMs. To reduce the computational burden of task similarity estimation, we employ bandit techniques to develop an algorithm based on lower confidence bounds to efficiently explore the task structure. Furthermore, we use sparse gradient updates to facilitate parameter optimization, making the approach better suited for LPMs. Theoretical analysis is provided to justify the rationale behind our approach, and experiments on both vision transformers (ViTs) and large language models (LLMs) demonstrate the effectiveness and efficiency of our approach across various domains, including vision and natural language processing tasks.

cs.LG↗

arXiv2025

Provably Efficient Online RLHF with One-Pass Reward Modeling

Reinforcement Learning from Human Feedback (RLHF) has shown remarkable success in aligning Large Language Models (LLMs) with human preferences. Traditional RLHF methods rely on a fixed dataset, which often suffers from limited coverage. To this end, online RLHF has emerged as a promising direction, enabling iterative data collection and refinement. Despite its potential, this paradigm faces a key bottleneck: the requirement to continuously integrate new data into the dataset and re-optimize the model from scratch at each iteration, resulting in computational and storage costs that grow linearly with the number of iterations. In this work, we address this challenge by proposing a one-pass reward modeling method that eliminates the need to store historical data and achieves constant-time updates per iteration. Specifically, we first formalize RLHF as a contextual preference bandit and develop a new algorithm based on online mirror descent with a tailored local norm, replacing the standard maximum likelihood estimation for reward modeling. We then apply it to various online RLHF settings, including passive data collection, active data collection, and deployment-time adaptation. We provide theoretical guarantees showing that our method enhances both statistical and computational efficiency. Finally, we design practical algorithms for LLMs and conduct experiments with the Llama-3-8B-Instruct and Qwen2.5-7B-Instruct models on Ultrafeedback and Mixture2 datasets, validating the effectiveness of our approach.

cs.LG↗