Institution profile

University of Liverpool

Academic institutioneurope · gb
Official website
Research library602linked papers
Opportunities0open roles
Selected work

Representative Papers

Tight Bounds for Quantum Phase Estimation and Related Problems

May 08, 2023Embedded Systems and Applications

This work establishes tight query complexity bounds—up to logarithmic factors—for quantum phase estimation (QPE) and its variants across all parameter regimes. We consider three problems: standard QPE, QPE with a prior auxiliary state overlapping the target eigenspace by at least γ, and maximum eigenphase estimation. Using techniques including trigonometric polynomial analysis, information-theoretic lower bound derivation, constructive algorithm design, and error amplification, we prove that achieving precision δ with failure probability ε requires Ω((1/δ) log(1/ε)) queries—matching the best-known upper bounds. Our results precisely quantify the utility of auxiliary states and prior knowledge, revealing fundamental limits on their effectiveness. Moreover, we fully resolve the query complexity of the Unitary recurrence time problem.

18 citations4 influentialRead paper

Task Adaptation from Skills: Information Geometry, Disentanglement, and New Objectives for Unsupervised Reinforcement Learning

Jun 12, 2025International Conference on Learning Representations

Unsupervised reinforcement learning (URL) aims to acquire transferable skills for unknown downstream tasks, yet existing mutual information-based skill learning (MISL) lacks theoretical characterization of skill transferability. We identify that skill diversity and separability are essential for efficient downstream policy initialization—properties not guaranteed by MISL. To address this, we propose two novel objectives—WSEP and PWSEP—grounded in Wasserstein geometry, along with a decoupling-aware metric, LSEPIN. Crucially, we establish the first theoretical link between Wasserstein distance and downstream adaptation cost, rigorously proving that our framework ensures complete discovery of optimal initial policies. Experiments demonstrate significant improvements in zero-shot transfer performance across multiple benchmarks, consistently outperforming MISL. Our method yields more disentangled skill representations and superior policy pretraining, enabling more effective downstream adaptation.

7 citationsRead paper

Jailbreaking and Mitigation of Vulnerabilities in Large Language Models

Oct 20, 2024arXiv.org

Existing research on large language models (LLMs) lacks a unified taxonomy for prompt injection and jailbreaking attacks, and insufficiently evaluates defenses under dynamic, interactive scenarios. Method: We propose the first four-dimensional attack taxonomy—spanning prompt-level, model-level, multimodal, and multilingual pathways—and develop a robust alignment framework tailored to interactive settings, alongside a novel automated jailbreaking detection method. We further conduct systematic defense benchmarking, bias diagnosis of existing evaluation benchmarks, and multi-dimensional security measurement. Contribution/Results: Our analysis reveals critical failure modes of current defenses in dynamic interactions, identifies key research gaps—including ethical implications and data bias—and delivers the first comprehensive technical roadmap for LLM safety alignment.

7 citationsRead paper

Deciding What is Good-for-MDPs

Feb 15, 2022International Conference on Concurrency Theory

This paper resolves the long-standing open problem of decidability for Good-for-MDPs (GFM) automata: given a nondeterministic Büchi automaton, determine whether it is GFM—i.e., whether it can reliably implement lightweight Büchi acceptance in MDP model checking and reinforcement learning. We establish the first decidability result for GFM, presenting an EXPTIME decision procedure and proving a matching PSPACE-hard lower bound, thereby settling EXPTIME-completeness. Our approach integrates game-theoretic modeling, simulation via alternating Turing machines, symbolic fixed-point computation, and parity game solving. This work fills a critical theoretical gap at the intersection of formal verification and reinforcement learning, providing a rigorous complexity characterization and algorithmic foundation for the construction, verification, and application of GFM automata.

6 citations1 influentialRead paper

Efficient and Stealthy Jailbreak Attacks via Adversarial Prompt Distillation from LLMs to SLMs

May 26, 2025arXiv.org

Jailbreaking large language models (LLMs) incurs high computational overhead and limited practicality due to reliance on costly LLM-based inference. Method: This paper proposes the first jailbreak-capability-oriented prompt distillation framework, transferring jailbreak knowledge efficiently from LLMs to small language models (SLMs). It integrates masked language modeling, reinforcement learning, and dynamic temperature control to construct a lightweight adversarial prompt generation and distillation pipeline. Contributions/Results: Experiments demonstrate high attack success rates across multiple mainstream LLMs, strong zero-shot cross-model transferability, and enhanced attack harmfulness. The approach reduces computational cost by 73% compared to LLM-based baselines, significantly improving stealthiness and real-world deployability of jailbreak attacks.

4 citationsRead paper
Recent publications

Latest Papers