arXiv subjects

Gong Yu

Publications and source records attributed to Gong Yu.

3 recordsLinked to original sources

arXiv2026

TokenPilot: Cache-Efficient Context Management for LLM Agents

As LLM agents are deployed in long-horizon sessions, context accumulation drives up inference costs. Existing approaches utilize text pruning or dynamic memory eviction to minimize token footprints; however, their unconstrained sequence mutations alter layouts, introducing prefix mismatches and cache invalidation. This reveals a critical trade-off between text sparsity and prompt cache continuity. To address this, we present TokenPilot, a dual-granularity context management framework. Globally, Ingestion-Aware Compaction acts as a framework harness to stabilize prompt prefixes and eliminate open-world environmental noise at the ingestion gate. Locally, Lifecycle-Aware Eviction monitors the ongoing residual utility of context segments, enforcing a conservative batch-turn schedule to offload content segments only when task relevance expires. Experiments on PinchBench and Claw-Eval under both isolated and continuous modes demonstrate that TokenPilot reduces costs by 61% and 56% in isolated mode, and 61% and 87% in continuous mode, while maintaining competitive performance compared to prior systems. TokenPilot has been integrated into LightRSI at https://github.com/zjunlp/RSI.

cs.CL↗

arXiv2025

RB-FT: Rationale-Bootstrapped Fine-Tuning for Video Classification

Vision Language Models (VLMs) are becoming increasingly integral to multimedia understanding; however, they often struggle with domain-specific video classification tasks, particularly in cases with limited data. This stems from a critical \textit{rationale gap}, where sparse domain data is insufficient to bridge the semantic distance between complex spatio-temporal content and abstract classification labels. We propose a two-stage self-improvement paradigm to bridge this gap without new annotations. First, we prompt the VLMs to generate detailed textual rationales for each video, compelling them to articulate the domain-specific logic. The VLM is then fine-tuned on these self-generated rationales, utilizing this intermediate supervision to align its representations with the nuances of the target domain. Second, conventional supervised fine-tuning (SFT) is performed on the task labels, achieving markedly higher effectiveness as a result of the model's pre-acquired domain reasoning. Extensive experiments on diverse datasets demonstrate that our method significantly outperforms direct SFT, validating self-generated rationale as an effective, annotation-efficient paradigm for adapting VLMs to domain-specific video analysis.

cs.CV↗

arXiv2017

Investigation of periodic-layered structure in Zn/CuxTiy systems

Diffusion couples Zn/CuxTiy were prepared by the melting contact method and then annealed at 663K for various times. Using scanning electron microscopy (SEM) with energy dispersive X-ray spectroscopy (EDS), we discovered 7 new systems, i.e. Zn/Cu9Ti, Zn/Cu4Ti, Zn/Cu2Ti, Zn/Cu7Ti3, Zn/Cu3Ti2, Zn/Cu11Ti9 and Zn/Cu9Ti11, which can form the interesting periodic-layered structure within the reaction zones. By the traditional metallographic treatment and the in-situ section observation method respectively, it was confirmed that the periodic-layered structure is really composed of the CuZn2 singer-phase and the (CuZn2+TiZn3) two-phase layers distributing alternately within the reaction front area. Furthermore, the thickness of the periodic layers relates to the composition of Cu-Ti substrates: the higher content of Cu atoms in C--Ti alloy substrates, the thinner the layers will be. The experimental results are in accordance to the prediction of the diffusion-induced stresses model.

cond-mat.mtrl-sci↗