arXiv subjects

Yifu Liu

Publications and source records attributed to Yifu Liu.

9 recordsLinked to original sources

arXiv2026

Explore More, Drift Less: Outcome-Only Reinforcement Learning Can Suffice for Long-Horizon Interactive Agents

Reinforcement learning is a natural way to post-train LLM agents for long-horizon interactive tasks judged only by end-of-task verification, yet a shared belief holds that outcome-only RL soon hits a ceiling on small open models. Recent work therefore compensates around the training with denser rewards, SFT priors, skill libraries, curated memory, or multi-agent orchestration. We argue the ceiling is an artifact of two failures of common practice. Signal starvation: group-relative RL with sparse outcome-only rewards yields a gradient only when a task's rollout group mixes successes and failures, so under-scaled exploration silences exactly the hardest, most instructive tasks. Policy drift: squeezing many updates out of a small task pool degrades the policy itself, as an unanchored objective lets the sampling distribution collapse exactly when saturation has already made informative groups rare. We present CANOPY (Coverage-ANchored On-PolicY RL), a minimalist protocol attacking both directly: scale same-task exploration until the natural signal reappears, keep every update on-policy, KL-anchored, and confined to the agent's own action tokens, then cash in an enlarged interaction budget at test time. On AppWorld, a long-horizon interactive coding benchmark, a Qwen3-14B policy trained with CANOPY through environment interaction alone--without task-specific supervision, auxiliary credit signals, or elaborate agent scaffolding--topped the public leaderboard (Feb. 2026; Test-Normal TGC 86.9, Test-Challenge 67.6), and the same design principles lift Qwen3.5-9B on SWE-bench Verified by 16.6 points. Agentic RL alone thus internalizes long-horizon capability directly into a small open model; we plan to release the complete training stack at https://github.com/AlibabaResearch/SignalCoverageRL.

cs.LG↗

arXiv2026

Affective Tools for Thought: Towards Shared Attention and Affective Reorienting in AI-Supported Thinking

Current Tools for Thought (TfTs) treat affect as either friction that slows cognitive progress or a signal to optimise it. Drawing on enactive cognitive science, we argue that affect is constitutive of cognition: it reshapes the trajectory of thinking, not just the speed. We identify two core barriers for Affective TfTs: the lack of Shared Attention (caring, directed attention to the user's mode of engagement) and the lack of Affective Reorienting (the capacity to use emotional moments to open new trajectories rather than reinforcing predetermined ones), and propose three design strategies that address both: Chain of Emotion X Chain of Thought, Affective Mirror, and Prompted Reorienting. The strategies are grounded in empirical findings from a study of a touch-aware conversational agent for embodied craft learning, and are oriented as provocations for future design.

cs.HC↗

arXiv2026

XiYan-SQL: A Novel Multi-Generator Framework For Text-to-SQL

To leverage the advantages of LLM in addressing challenges in the Text-to-SQL task, we present XiYan-SQL, an innovative framework effectively generating and utilizing multiple SQL candidates. It consists of three components: 1) a Schema Filter module filtering and obtaining multiple relevant schemas; 2) a multi-generator ensemble approach generating multiple highquality and diverse SQL queries; 3) a selection model with a candidate reorganization strategy implemented to obtain the optimal SQL query. Specifically, for the multi-generator ensemble, we employ a multi-task fine-tuning strategy to enhance the capabilities of SQL generation models for the intrinsic alignment between SQL and text, and construct multiple generation models with distinct generation styles by fine-tuning across different SQL formats. The experimental results and comprehensive analysis demonstrate the effectiveness and robustness of our framework. Overall, XiYan-SQL achieves a new SOTA performance of 75.63% on the notable BIRD benchmark, surpassing all previous methods. It also attains SOTA performance on the Spider test set with an accuracy of 89.65%.

cs.CL↗

arXiv2026

ProVoice: Designing Proactive Functionality for In-Vehicle Conversational Assistants using Multi-Objective Bayesian Optimization to Enhance Driver Experience

The next step for In-vehicle Conversational Assistants (IVCAs) will be their capability to initiate and automate proactive system interactions throughout journeys. However, diverse drivers make it challenging to design voice interventions tailored towards individual on-road expectations. This paper evaluates the effectiveness of Human-in-the-Loop (HITL) Multi-Objective Bayesian Optimization (MOBO) in design by implementing ProVoice: a Virtual Reality (VR) driving simulator integrating MOBO to investigate the effects of IVCA design variants on perceived mental demand, predictability, and usefulness. By reporting the Pareto Front from a within-subjects VR study (N=19), this paper proposes optimal design trade-offs. Follow-up analysis demonstrates MOBO's success in discovering effective intervention strategies, with reduced participant mental demand, alongside enhanced predictability and usefulness while engaging with the proactive IVCA. Implications for computational techniques in future research on proactive intervention strategies are discussed. ProVoice can extend to include alternative design parameters and driving scenarios, encouraging intervention design on a broad scale.

cs.HC↗

arXiv2025

A Preview of XiYan-SQL: A Multi-Generator Ensemble Framework for Text-to-SQL

To tackle the challenges of large language model performance in natural language to SQL tasks, we introduce XiYan-SQL, an innovative framework that employs a multi-generator ensemble strategy to improve candidate generation. We introduce M-Schema, a semi-structured schema representation method designed to enhance the understanding of database structures. To enhance the quality and diversity of generated candidate SQL queries, XiYan-SQL integrates the significant potential of in-context learning (ICL) with the precise control of supervised fine-tuning. On one hand, we propose a series of training strategies to fine-tune models to generate high-quality candidates with diverse preferences. On the other hand, we implement the ICL approach with an example selection method based on named entity recognition to prevent overemphasis on entities. The refiner optimizes each candidate by correcting logical or syntactical errors. To address the challenge of identifying the best candidate, we fine-tune a selection model to distinguish nuances of candidate SQL queries. The experimental results on multiple dialect datasets demonstrate the robustness of XiYan-SQL in addressing challenges across different scenarios. Overall, our proposed XiYan-SQL achieves the state-of-the-art execution accuracy of 75.63% on Bird benchmark, 89.65% on the Spider test set, 69.86% on SQL-Eval, 41.20% on NL2GQL. The proposed framework not only enhances the quality and diversity of SQL queries but also outperforms previous methods.

cs.AI↗

arXiv2024

MoMQ: Mixture-of-Experts Enhances Multi-Dialect Query Generation across Relational and Non-Relational Databases

The improvement in translating natural language to structured query language (SQL) can be attributed to the advancements in large language models (LLMs). Open-source LLMs, tailored for specific database dialects such as MySQL, have shown great performance. However, cloud service providers are looking for a unified database manager service (e.g., Cosmos DB from Azure, Amazon Aurora from AWS, Lindorm from AlibabaCloud) that can support multiple dialects. This requirement has led to the concept of multi-dialect query generation, which presents challenges to LLMs. These challenges include syntactic differences among dialects and imbalanced data distribution across multiple dialects. To tackle these challenges, we propose MoMQ, a novel Mixture-of-Experts-based multi-dialect query generation framework across both relational and non-relational databases. MoMQ employs a dialect expert group for each dialect and a multi-level routing strategy to handle dialect-specific knowledge, reducing interference during query generation. Additionally, a shared expert group is introduced to address data imbalance, facilitating the transfer of common knowledge from high-resource dialects to low-resource ones. Furthermore, we have developed a high-quality multi-dialect query generation benchmark that covers relational and non-relational databases such as MySQL, PostgreSQL, Cypher for Neo4j, and nGQL for NebulaGraph. Extensive experiments have shown that MoMQ performs effectively and robustly even in resource-imbalanced scenarios.

cs.CL↗

arXiv2023

Towards Loose-Fitting Garment Animation via Generative Model of Deformation Decomposition

Existing data-driven methods for garment animation, usually driven by linear skinning, although effective on tight garments, do not handle loose-fitting garments with complex deformations well. To address these limitations, we develop a garment generative model based on deformation decomposition to efficiently simulate loose garment deformation without directly using linear skinning. Specifically, we learn a garment generative space with the proposed generative model, where we decouple the latent representation into unposed deformed garments and dynamic offsets during the decoding stage. With explicit garment deformations decomposition, our generative model is able to generate complex pose-driven deformations on canonical garment shapes. Furthermore, we learn to transfer the body motions and previous state of the garment to the latent space to regenerate dynamic results. In addition, we introduce a detail enhancement module in an adversarial training setup to learn high-frequency wrinkles. We demonstrate our method outperforms state-of-the-art data-driven alternatives through extensive experiments and show qualitative and quantitative analysis of results.

cs.CV↗

arXiv2023

JCDNet: Joint of Common and Definite phases Network for Weakly Supervised Temporal Action Localization

Weakly-supervised temporal action localization aims to localize action instances in untrimmed videos with only video-level supervision. We witness that different actions record common phases, e.g., the run-up in the HighJump and LongJump. These different actions are defined as conjoint actions, whose rest parts are definite phases, e.g., leaping over the bar in a HighJump. Compared with the common phases, the definite phases are more easily localized in existing researches. Most of them formulate this task as a Multiple Instance Learning paradigm, in which the common phases are tended to be confused with the background, and affect the localization completeness of the conjoint actions. To tackle this challenge, we propose a Joint of Common and Definite phases Network (JCDNet) by improving feature discriminability of the conjoint actions. Specifically, we design a Class-Aware Discriminative module to enhance the contribution of the common phases in classification by the guidance of the coarse definite-phase features. Besides, we introduce a temporal attention module to learn robust action-ness scores via modeling temporal dependencies, distinguishing the common phases from the background. Extensive experiments on three datasets (THUMOS14, ActivityNetv1.2, and a conjoint-action subset) demonstrate that JCDNet achieves competitive performance against the state-of-the-art methods. Keywords: weakly-supervised learning, temporal action localization, conjoint action

cs.CV↗

arXiv2021

DCANet: Dense Context-Aware Network for Semantic Segmentation

As the superiority of context information gradually manifests in advanced semantic segmentation, learning to capture the compact context relationship can help to understand the complex scenes. In contrast to some previous works utilizing the multi-scale context fusion, we propose a novel module, named Dense Context-Aware (DCA) module, to adaptively integrate local detail information with global dependencies. Driven by the contextual relationship, the DCA module can better achieve the aggregation of context information to generate more powerful features. Furthermore, we deliberately design two extended structures based on the DCA modules to further capture the long-range contextual dependency information. By combining the DCA modules in cascade or parallel, our networks use a progressive strategy to improve multi-scale feature representations for robust segmentation. We empirically demonstrate the promising performance of our approach (DCANet) with extensive experiments on three challenging datasets, including PASCAL VOC 2012, Cityscapes, and ADE20K.

cs.CV↗