Author
Harbin Institute of Technology (Shenzhen)
We have 9 of 51 papers
We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.
Not the right person? Other researchers publish under this name.
By avoiding weight updates entirely, Nullify serves as an efficient, plug-and-play inference-time intervention framework for LLM unlearning, and matches or surpasses established baselines in forget quality while achieving near-lossless preservation of model utility.
Large Language Models (LLMs) have achieved remarkable capabilities but remain vulnerable to jailbreak attacks that elicit harmful or unsafe outputs. Existing safety alignment approaches, including Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF), often require substantial attack-specif...
HEAR is established as a reusable coordination substrate for efficient agentic LLM serving by separating protocol semantics from optimization policies, and Mooncake shows that workflow intent and live engine state provide complementary benefits across load regimes.
Long-context LLM agents accumulate interaction histories that strain KV-cache memory and attention computation. Although sparse attention reduces these costs, heterogeneous cache representations and workflows hinder integration with existing inference engines, while prior sparse-serving abstractions support only specif...
ResidualKV is proposed, which factorizes the KV cache into a sparse set of globally retrieved references and compact, quantized residual codes for the remaining tokens, and shows that global cross-token redundancy supports accurate, memory-efficient, and computation-efficient long-context inference.
IDRBench is introduced, a benchmark for evaluating interactive deep research with controlled opportunities for clarification, and shows that access to clarification alone does not guarantee better outcomes: success depends on what agents ask and how effectively they incorporate the resulting feedback.
A new TSE setting for continuous, magnitude-aware condition transitions is introduced and JAVELIN, a retrieval-guided framework for directional editing via JAcobian-VEctor Latent INference is proposed, enabling precise, content-preserving edits without retraining the generative model.
ShallowStream is a novel framework that leverages the shallow layers of an MLLM to simultaneously perform frame encoding and retrieval index building and achieves performance on par with the strongest existing streaming methods, while reducing per-frame prefill latency and 10-second end-to-end latency by up to 52.1x an...
Low-Rank Clone (LRC), an efficient pre-training method that constructs SLMs aspiring to behavioral equivalence with strong teacher models, and maximizes knowledge transfer while removing the need for explicit alignment modules.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
