arXiv subjects

Kai Sheng

Publications and source records attributed to Kai Sheng.

11 recordsLinked to original sources

arXiv2026

KuaiRP Series Role-playing Models Technical Report

This paper introduces the complete technical solution for the KuaiRP series of role-playing models. We aim to achieve four core objectives for a dedicated role-playing model: simplified prompt engineering, highly stable output quality, built-in domain world knowledge, and high-efficiency deployment with a small parameter size. However, effectively injecting deep domain knowledge often leads to a severe catastrophic forgetting of the model's general agent capabilities. To overcome this trade-off, we propose a multi-stage training pipeline. First, we design a standardized character template and construct an SFT data pipeline based on user behavior simulation and reverse profile filtering. Next, we utilize a rule-based composite reward function during the Reinforcement Learning (RL) phase to eliminate common degradation phenomena like length expansion and repetitive generation. Finally, to recover the general capabilities compromised during SFT and RL, we propose a novel self-distillation paradigm using Two-stage On-Policy Distillation (OPD) equipped with Cumulative-Divergence Decay (CDD). By using the domain-adapted model as the teacher and the original base model as the student, we effectively balance deep domain knowledge injection with the preservation of general agent capabilities. Experimental results demonstrate that the KuaiRP models not only match the current state-of-the-art proprietary models in role-playing fidelity within our target domains, but also successfully recover general agent capabilities, maintaining extremely low deployment costs.

cs.AI↗

arXiv2026

A partial answer to Brezis' Open Problem 2.1

In this paper, we give a partial answer to Brezis' Open Problem~2.1, which concerns the uniqueness of solutions to the Ginzburg--Landau equation in the unit disc with the degree-one boundary condition. Let $\lambda_1$ be the first Dirichlet eigenvalue of $-\Delta$ in the unit disc and set $\varepsilon_*:=\lambda_1^{-1/2}$. We prove that there exists $\delta>0$ such that the radial solution is the unique weak solution for every $\varepsilon\in(\varepsilon_*-\delta,\infty)$. More generally, we establish the analogous uniqueness result for the Ginzburg--Landau system on bounded connected $C^{1,1}$ domains in $\mathbb R^N$, $2\leq N\leq4$, with nontrivial boundary data. In particular, uniqueness persists slightly below the convexity threshold $\varepsilon_*$, where the strict convexity argument is no longer available. The proof combines strict convexity for $\varepsilon\geq\varepsilon_*$ with a compactness argument, nondegeneracy of the solution at $\varepsilon=\varepsilon_*$ and the implicit function theorem.

math.AP↗

arXiv2026

TRACE Bench: Task-driven Roleplay Agentic Checklist Evaluation

Roleplay evaluation should do more than assign a single score: it should reveal which role requirements were tested, which failed, and which dialogue evidence supports the judgment. We propose TRACE Bench, a task-driven agentic checklist evaluation framework. It decomposes each role profile offline into a fixed checklist, then uses a User Agent to converse naturally with the target roleplay model while privately updating checklist states from model responses. Scores therefore trace back to checklist items and supporting dialogue turns rather than a black-box holistic impression. For coverage cross-validation, we audit released M2 free-dialogue transcripts from the MiniMax Role-play Benchmark against the same role-derived checklist. The released free-chat transcripts cover only 73.74% of key role-profile points, whereas TRACE Bench reaches 99.91% coverage in fewer turns. Robustness experiments show stable rankings under repeated runs and User Agent replacement. Across 26 models, TRACE Bench reports overall rankings together with capability breakdowns and checklist traces. It also supports Closed-Loop Benchmark Evolution, distilling verification methods proven effective in failed traces so later evaluations can more reliably elicit and examine observed failure modes.

cs.CL↗

arXiv2026

A Comprehensive Survey and Systematic Real-World Evaluation of Embodied Vision-and-Language Navigation

Navigation is a fundamental capability of autonomous systems, yet most existing approaches rely on highly structured models and strong prior assumptions, limiting their robustness in open and uncertain real-world environments. Vision-and-Language Navigation (VLN) offers a promising direction by enabling robots to integrate natural language understanding with visual perception in a data-driven manner. Although VLN has attracted increasing research attention, systematic methodological taxonomy and real-world validation remain limited. This survey presents a comprehensive review of VLN research. Specifically, state-of-the-art methods are organized along two orthogonal dimensions: action paradigms, including hierarchical and monolithic frameworks, and model paradigms, including discriminative and generative approaches. A critical analysis of their respective strengths and limitations is provided. Additionally, we conduct a systematic real-world evaluation of representative VLN system configurations on a physical robotic platform. Experiments across ten diverse real-world scenes show a substantial performance gap between simulation and real-world deployment under the tested configurations: a representative monolithic RGB-only method achieves 61% success in simulation but drops to 22% in real-world deployment, while a hierarchical framework achieves a higher real-world success rate of 51%, suggesting stronger robustness in our evaluation setting. Finally, we highlight key challenges in perception, decision-making, and control that must be addressed in future research.

cs.RO↗

arXiv2026

FAST: A Holistic Framework for Optimizing Memory-I/O, Computation, and Sampling in Temporal GNN Training

Temporal Graph Neural Networks (TGNNs) are widely used for learning from dynamic graphs in applications such as recommendation, social network analysis, and traffic forecasting. However, scaling TGNN training to large dynamic graphs remains challenging due to three intertwined bottlenecks: memory I/O, irregular computation, and temporal neighbor sampling. Existing systems often optimize these stages in isolation, leaving substantial performance headroom on the table. We present FAST, a holistic framework that accelerates end-to-end TGNN training by jointly optimizing sampling, memory I/O, and computation. FAST introduces SlimCache, which exploits within-batch compression and cross-batch caching to reduce host-device data movement under limited GPU memory budgets. It further designs thread-efficient graph operators tailored to sparse temporal subgraphs, improving GPU cache locality and reducing the latency of aggregation and edge softmax. In addition, FAST employs a topology-aware sampling strategy that improves CPU cache locality and accelerates temporal neighbor sampling. Extensive experiments on real-world large dynamic graphs show that FAST achieves an average of 2.1x (up to 4.7x) speedup over state-of-the-art systems without sacrificing model accuracy.

cs.LG↗

arXiv2026

P2DNav: Panorama-to-Downview Reasoning for Zero-shot Vision-and-Language Navigation

Vision-and-language navigation (VLN) requires an embodied agent to ground natural-language instructions into executable navigation actions in unseen environments. Existing zero-shot methods typically rely on additional waypoint prediction modules, which often entangle high-level directional reasoning with fine-grained local grounding, leading to error-prone and unstable decisions. In this paper, we propose P2DNav, a hierarchical framework for zero-shot vision-and-language navigation. P2DNav consists of three core components: Panorama-to-Downview (P2D), Sliding-Window Dialogue Memory (SDM), and Reflective Reorientation Mechanism (RRM). P2D explicitly decomposes navigation decision-making into two stages: panoramic direction selection and downview local grounding. It first selects the instruction-relevant direction from a 360{\deg} panorama, and then predicts a pixel-level target point from the downview RGB observation in that direction. In addition, SDM organizes navigation history as a multi-turn dialogue context and maintains recent visual observations within a sliding window to support long-horizon navigation. RRM further enables reflective reorientation by assessing the reliability of local grounding based on the downview observation and returning to panoramic direction selection when necessary. Experiments on the R2R-CE benchmark show that P2DNav achieves strong performance among zero-shot methods. In particular, compared with the state-of-the-art (SOTA) zero-shot waypoint-based and waypoint-free methods, P2DNav achieves SR gains of 146.6% and 58.9%, respectively, demonstrating the effectiveness of P2D, SDM, and RRM for zero-shot VLN. Code will be released for public use.

cs.CV↗

arXiv2026

Cutscene Agent: An LLM Agent Framework for Automated 3D Cutscene Generation

Cutscenes are carefully choreographed cinematic sequences embedded in video games and interactive media, serving as the primary vehicle for narrative delivery, character development, and emotional engagement. Producing cutscenes is inherently complex: it demands seamless coordination across screenwriting, cinematography, character animation, voice acting, and technical direction, often requiring days to weeks of collaborative effort from multidisciplinary teams to produce minutes of polished content. In this work, we present Cutscene Agent, an LLM agent framework for automated end-to-end cutscene generation. The framework makes three contributions: (1)~a Cutscene Toolkit built on the Model Context Protocol (MCP) that establishes \emph{bidirectional} integration between LLM agents and the game engine -- agents not only invoke engine operations but continuously observe real-time scene state, enabling closed-loop generation of editable engine-native cinematic assets; (2)~a multi-agent system where a director agent orchestrates specialist subagents for animation, cinematography, and sound design, augmented by a visual reasoning feedback loop for perception-driven refinement; and (3)~CutsceneBench, a hierarchical evaluation benchmark for cutscene generation. Unlike typical tool-use benchmarks that evaluate short, isolated function calls, cutscene generation requires long-horizon, multi-step orchestration of dozens of interdependent tool invocations with strict ordering constraints -- a capability dimension that existing benchmarks do not cover. We evaluate a range of LLMs on CutsceneBench and analyze their performance across this challenging task.

cs.GR↗

arXiv2025

Normalized Solutions for Schr\"odinger-Bopp-Podolsky Systems in Bounded Domains with General Nonlinearities

In this paper, by adapting the perturbation method, we study normalized standing wave solutions for the following nonlinear Schr\"odinger-Bopp-Podolsky system: - Delta u + q(x) phi u = omega u + f(u) in Omega, - Delta phi + a^2 Delta^2 phi = q(x) u^2 in Omega, where Omega is a smooth bounded domain in R^3, a > 0, and omega is the Lagrange multiplier associated with the L^2 mass constraint integral over Omega of u^2 equals mu, and f: R -> R is a continuous function satisfying some technical conditions. We introduce a perturbation framework for the problem and investigate normalized solutions. In particular, we prove the existence of normalized solutions for all masses mu in an interval (0, mu_0), under either Navier or Neumann boundary conditions for phi. Moreover, when f is odd, we obtain multiplicity of normalized solutions; and if Omega is star-shaped, we further obtain a normalized ground state solution.

math.AP↗

arXiv2025

FlexQ: Efficient Post-training INT6 Quantization for LLM Serving via Algorithm-System Co-Design

Large Language Models (LLMs) demonstrate exceptional performance but entail significant memory and computational costs, restricting their practical deployment. While existing INT4/INT8 quantization reduces these costs, they often degrade accuracy or lack optimal efficiency. INT6 quantization offers a superior trade-off between model accuracy and inference efficiency, but lacks hardware support in modern GPUs, forcing emulation via higher-precision arithmetic units that limit acceleration. In this paper, we propose FlexQ, a novel post-training INT6 quantization framework combining algorithmic innovation with system-level optimizations. FlexQ employs uniform 6-bit weight quantization across all layers, with adaptive retention of 8-bit activations in layers identified through layer-wise sensitivity analysis. To maximize hardware efficiency, we develop a specialized high-performance GPU kernel supporting matrix multiplication for W6A6 and W6A8 representations via Binary Tensor Core (BTC) equivalents, effectively bypassing the lack of native INT6 tensor cores. Evaluations on LLaMA family models show FlexQ maintains near-FP16 accuracy, with perplexity increases of no more than 0.1 on WikiText2. The proposed kernel achieves an average 1.39$\times$ speedup over ABQ-LLM on LLaMA-2-70B linear layers. End-to-end, FlexQ delivers 1.33$\times$ inference acceleration and 1.21$\times$ memory savings over SmoothQuant. Code is released at https://github.com/FlyFoxPlayer/FlexQ.

cs.LG↗

arXiv2025

CARD: A Cache-Assisted Parallel Speculative Decoding Framework via Query-and-Correct Paradigm for Accelerating LLM Inference

Speculative decoding (SD), where a draft model provides multiple candidate tokens for the target model to verify in parallel, has demonstrated significant potential for accelerating LLM inference. Yet, existing SD approaches adhere to a strict draft-then-verify paradigm, enforcing a sequential process that hampers performance and constrains the draft model's capacity. Moreover, rejecting a token in the candidate sequence invalidates all subsequent tokens, leading to wasted computation during drafting. To overcome these limitations, we propose a cache-assisted parallel speculative decoding framework called CARD, which employs a novel query-and-correct paradigm. Our approach decouples drafting from verification: the draft model populates a shared cache with candidate tokens, while the target model concurrently refines the draft's trajectory. This enables inference at near-draft-speed, effectively leveraging the draft model's efficiency without additional fine-tuning. Experimental results show that CARD significantly outperforms existing state-of-the-art methods, achieving up to a 4.83x acceleration over vanilla autoregressive decoding, with no fine-tuning required for either models.

cs.LG↗

arXiv2025

ILNet: Trajectory Prediction with Inverse Learning Attention for Enhancing Intention Capture

Trajectory prediction for multi-agent interaction scenarios is a crucial challenge. Most advanced methods model agent interactions by efficiently factorized attention based on the temporal and agent axes. However, this static and foward modeling lacks explicit interactive spatio-temporal coordination, capturing only obvious and immediate behavioral intentions. Alternatively, the modern trajectory prediction framework refines the successive predictions by a fixed-anchor selection strategy, which is difficult to adapt in different future environments. It is acknowledged that human drivers dynamically adjust initial driving decisions based on further assumptions about the intentions of surrounding vehicles. Motivated by human driving behaviors, this paper proposes ILNet, a multi-agent trajectory prediction method with Inverse Learning (IL) attention and Dynamic Anchor Selection (DAS) module. IL Attention employs an inverse learning paradigm to model interactions at neighboring moments, introducing proposed intentions to dynamically encode the spatio-temporal coordination of interactions, thereby enhancing the model's ability to capture complex interaction patterns. Then, the learnable DAS module is proposed to extract multiple trajectory change keypoints as anchors in parallel with almost no increase in parameters. Experimental results show that the ILNet achieves state-of-the-art performance on the INTERACTION and Argoverse motion forecasting datasets. Particularly, in challenged interaction scenarios, ILNet achieves higher accuracy and more multimodal distributions of trajectories over fewer parameters. Our codes are available at https://github.com/mjZeng11/ILNet.

cs.CV↗