arXiv subjects

Yichuan Liu

Publications and source records attributed to Yichuan Liu.

4 recordsLinked to original sources

arXiv2026

OUTLETS: Output-Length Prediction from Speculative Decoding Backbones

The heavy-tailed distribution of output lengths in Large Language Model (LLM) serving poses major challenges for resource provisioning and cluster scheduling. Although output-length prediction can mitigate these issues, existing approaches have key drawbacks: external proxy models add substantial latency and often have limited fidelity, whereas internal state-based methods are efficient but rely on shallow probes of current model states. We identify a structural connection between speculative decoding (SD) and length prediction: latent representations produced by the draft decoder in advanced frameworks (e.g., EAGLE-3) encode signals that are predictive of generation length. Building on this insight, we introduce OUTLETS (Output-Length Prediction from Speculative Decoding Backbones), which repurposes the speculative backbone as a trajectory-aware length predictor. When its draft representations are already computed for speculative decoding, OUTLETS adds only a lightweight regression head and achieves lower MAE than the evaluated methods. Under saturated disaggregated serving, OUTLETS predictions enable standard scheduling policies to prioritize shorter requests and distribute requests more evenly across decoding instances, reducing short-request P99 latency by 34.8%.

cs.CL↗

arXiv2026

WaveformQA: Benchmarking LLM Temporal Reasoning on Digital Waveforms

Large Language Models (LLMs) have demonstrated strong capabilities in code generation and reasoning, yet their ability to perform temporal reasoning over digital waveform data remains largely unexplored. Although reasoning over digital waveforms is a critical bottleneck in design verification, existing benchmarks primarily evaluate hardware description language (HDL) code generation and use waveforms only as supplementary context. This paper presents WaveformQA, an open-source question-answering benchmark for evaluating LLM temporal reasoning over digital waveforms. The benchmark comprises 360 questions with programmatically generated ground truths across eight categories of varying difficulty, including questions targeting multi-signal correlation and event ordering. Waveforms are generated from open-source design implementations, ensuring reproducibility and grounding the benchmark in real hardware behavior. Evaluation of frontier LLMs reveals that while models achieve reasonable accuracy on simple queries, performance degrades due to context window limitations and reasoning difficulties on complex temporal and multi-step questions. In addition, we show that an event-time JSON representation of waveforms improves LLM reasoning accuracy versus the standardized value change dump (VCD) format. The open-source framework supports extending to new question categories and importing new waveform sources, enabling researchers to rapidly prototype temporal reasoning experiments.

cs.AI↗

arXiv2025

A Model-Based Derivative-Free Optimization Algorithm for Partially Separable Problems

We propose UPOQA, a derivative-free optimization algorithm for partially separable unconstrained problems, leveraging quadratic interpolation and a structured trust-region framework. By decomposing the objective into element functions, UPOQA constructs underdetermined element models and solves subproblems efficiently via a modified projected gradient method. Innovations include an approximate projection operator for structured trust regions, improved management of elemental radii and models, a starting point search mechanism, and support for hybrid black-white-box optimization, etc. Numerical experiments on 85 CUTEst problems demonstrate that \texttt{UPOQA} can significantly reduce the number of function evaluations. To quantify the impact of exploiting partial separability, we introduce the speed-up profile to further evaluate the acceleration effect. Results show that the speed-up of UPOQA over baselines is less significant in low-precision scenarios but becomes more pronounced in high-precision scenarios. Applications to quantum variational problems further validate its practical utility.

math.OC↗

arXiv2021

MRZ code extraction from visa and passport documents using convolutional neural networks

Detecting and extracting information from Machine-Readable Zone (MRZ) on passports and visas is becoming increasingly important for verifying document authenticity. However, computer vision methods for performing similar tasks, such as optical character recognition (OCR), fail to extract the MRZ given digital images of passports with reasonable accuracy. We present a specially designed model based on convolutional neural networks that is able to successfully extract MRZ information from digital images of passports of arbitrary orientation and size. Our model achieved 100% MRZ detection rate and 98.36% character recognition macro-f1 score on a passport and visa dataset.

cs.CV↗