arXiv subjects

Zezhong Wang

Publications and source records attributed to Zezhong Wang.

At least 19 recordsLinked to original sources

arXiv2026

Write, Execute, Refine: From Skill Followers to Skill Optimizers via Reinforcement Learning from Execution Feedback

Expert-written natural language skills can improve tool-using agents, yet agent-authored skills perform 8-11 points worse than using no skill. This gap suggests that following procedural guidance and improving it from execution evidence are distinct capabilities. Inference time loops can repair skills but do not improve the model that writes the next one. We study how to organize execution experience from intermediate skills into training states for an optimizer. We introduce WER (Write, Execute, and Refine), a multi-phase framework that trains a Skill Optimizer outside a frozen executor. The optimizer proposes skills, a frozen agent executes each repeatedly, and a programmatic verifier scores the outcomes. The scores provide relative credit and select mixed-outcome records. Matched successful and failed trajectories from these records form the next phase's refinement states, so the optimizer learns from the consequences of its earlier outputs. On BFCL v4 multi-turn and tau2-bench, WER improves average Pass@1 over the no-skill baseline by 7.80 and 3.85 points, respectively. Under an identical refinement workflow, it outperforms the same backbone without optimizer training by 9.35 and 10.29 points. The trained 4B optimizer reaches 76.63 percent on BFCL v4, outperforming all evaluated off-the-shelf general-purpose models used as skill optimizers on average.

cs.CL↗

arXiv2026

Epistemic Tensions: Reframing A Visualization Co-Design through Entanglement Theory

In this work, we present how employing the lens of entanglement helped us examine and reframe epistemic tensions arising in a visualization co-design project. Entanglement theory challenges traditional assumptions in the visualization research community by emphasizing that knowledge is not produced through linear, isolated processes, but is inherently entangled with phenomena and apparatuses. While this perspective offers a compelling critique of conventional research practices, its practical value for visualization research remains underexplored. We apply the entanglement lens to examine and reframe the epistemic tensions that emerged in a longitudinal community-based visualization co-design project. Our experience shows that the entanglement perspective not only provides a richer understanding of these tensions, but also helps transform them into generative opportunities for methodological and theoretical reflection. Applying this lens enabled us to critically interrogate the language used in research, to develop a more nuanced understanding of visualization co-design, and to surface ``dark sides'' of conventional visualization design pipelines. These contributions illustrate the practical value of embracing entanglement as an epistemological lens for visualization research.

cs.HC↗

arXiv2026

Looped World Models

Current world models face a fundamental tension: faithful long-horizon simulation demands deep computation, but deeper models are expensive to deploy and prone to compounding errors. We resolve this by introducing Looped World Models (LoopWM), which are the first looped architectures for world modelling. Our method iteratively refines latent environment states through a parameter-shared transformer block. This yield up to 100x parameter efficiency over conventional approaches with adaptive computation that automatically scales depth to match the complexity of each prediction step. Orthogonal to scaling model size and training data, LoopWM establishes iterative latent depth as a new scaling axis for world simulation, which might significantly push the community forward.

cs.LG↗

arXiv2026

Beyond the Literal: Decomposing Pragmatic Intent in Multimodal Meme Understanding

When asked what a meme or sarcastic post means, Large Vision Language Models (LVLMs) tend to describe what the image shows rather than what the author is trying to communicate. Standard instruction tuning entangles a post's literal content with its pragmatic meaning, letting surface-level details contaminate the final response. We reframe meme understanding as a problem of literal-pragmatic decomposition and propose \textbf{Intent Projection}, a framework that separates the two signals at the representation, output, and objective levels within a single LVLM backbone. At the representation level, an orthogonal projection module removes dominant unimodal directions from the fused image-text representation, retaining only the pragmatic residual, while a surface-real affect classifier anchors the decoder with a discrete tag that names the polarity gap. At the output level, the model externalizes a structured reasoning chain, and at the objective level a contrastive reward explicitly penalizes answers that restate the literal description. Across six multimodal benchmarks, Intent Projection consistently outperforms open-source baselines and narrows the gap to proprietary models, with the largest gains on high-divergence posts where literal collapse is most damaging.

cs.CL↗

arXiv2026

Bridging Auxiliary Constraints to Resolve Instruction Following in Large Reasoning Models

Large Reasoning Models (LRMs) have demonstrated impressive capabilities in many tasks, yet they struggle with reliably following multiple instructions, either by failing to satisfy individual constraints or by struggling to balance competing constraints simultaneously. We formalize this challenge as the Constraint Adherence Problem (CAP). This paper introduces a novel framework that addresses CAP by representing instructions as a structured knowledge graph of constraints. Our approach, Constraint Relationship Graph Completion (CRGC), explicitly models relationships between constraints, identifies adherence challenges, and discovers ``bridge constraints'' that help the model better focus on and reconcile requirements. Bridge constraints act as auxiliary instructions that make primary constraints more salient and compatible. Unlike existing approaches that enhance instruction following through general training methods, CRGC specifically improves constraint satisfaction by leveraging the model's own knowledge to create better pathways for generation. Experiments across three popular instruction following datasets demonstrate that our approach reduces constraint violations by 39% compared to standard prompting while maintaining reasoning abilities of large reasoning models.

cs.AI↗

arXiv2026

Guaranteeing Knowledge Integration with Joint Decoding for Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) significantly enhances Large Language Models (LLMs) by providing access to external knowledge. However, current research primarily focuses on retrieval quality, often overlooking the critical ''integration bottleneck'': even when relevant documents are retrieved, LLMs frequently fail to utilize them effectively due to conflicts with their internal parametric knowledge. In this paper, we argue that implicitly resolving this conflict in a single generation pass is suboptimal. We introduce GuarantRAG, a framework that explicitly decouples reasoning from evidence integration. First, we generate an ''Inner-Answer'' based solely on parametric knowledge to capture the model's reasoning flow. Second, to guarantee faithful evidence extraction, we generate a ''Refer-Answer'' using a novel Contrastive DPO objective. This objective treats the parametric Inner-Answer as a negative constraint and the retrieved documents as positive ground truth, forcing the model to suppress internal hallucinations in favor of external evidence during this phase. Finally, rather than naive concatenation or using the DPO trained model directly, we propose a joint decoding mechanism that dynamically fuses the logical coherence of the Inner-Answer with the factual precision of the Refer-Answer at the token level. Experiments on five QA benchmarks demonstrate that GuarantRAG improves accuracy by up to 12.1% and reduces hallucinations by 16.3% compared to standard and dynamic RAG baselines.

cs.CL↗

arXiv2026

The Fuzzy Front Ends: Reflections on the Never-Ending Story of Visualization Co-Design

Co-design is an increasingly popular approach in HCI and visualization, yet there is little guidance on how to effectively apply this method in visualization contexts. In this paper, we visually present our experience of a two-and-a-half-year co-design project with the local arts community. Focusing on facilitating community exploration and sense-making around arts funding distribution, the project involved a series of co-design sessions between visualization researchers and members of the arts community. Through these iterative sessions, we built shared understanding and developed visualization prototypes tailored to community needs. However, the practice is far from complete, and we found ourselves continually returning to the "fuzzy front end" of the co-design process. We share this ongoing story through comic-style visuals and reflect on three fuzzy front ends that we encountered during the project. By sharing these experiences with the visualization community, we hope to offer insights that others can draw on in their own community-engaged co-design work.

cs.HC↗

arXiv2026

Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors

Large language models (LLMs) can call tools effectively, yet they remain brittle in multi-turn execution: after a tool-call error, smaller models often fall into repetitive invalid re-invocations instead of interpreting the feedback and recovering. This failure mode persists because current training paradigms do not explicitly teach models how to recover from execution errors. In particular, standard reinforcement learning (RL) collapses rich failure experience into sparse negative rewards, while pre-collected error-correction datasets become mismatched to the policy's evolving failure modes. To bridge this gap, we propose Fission-GRPO, a framework that converts execution errors into on-policy corrective supervision within the RL training loop. Our core mechanism fissions each failed trajectory into a new training instance by augmenting it with diagnostic feedback from a fine-tuned Error Simulator, then resampling multiple recovery rollouts on-policy. This enables the model to learn from the precise errors it makes during exploration, rather than from static, pre-collected error cases. On BFCL v4 Multi-Turn, Fission-GRPO improves the error recovery rate of Qwen3-8B by 5.7% absolute and overall accuracy by 4.0% (from 42.75% to 46.75%), outperforming both RL baselines and specialized tool-use agents. The method further generalizes to TAU-Bench and TAU2-Bench, achieving leading results across most settings with gains up to +17.4%.

cs.LG↗

arXiv2025

What Do We Mean When We Talk About Data Storytelling?

Data storytelling has seen rapid growth through a proliferation of examples, as well as theoretical and technical advancements contributed across multiple disciplines. In this paper, we present a comprehensive survey of data storytelling research from 2010 to 2025. By analyzing the conceptualizations of data storytelling collected from related publications, we reveal the field's perspectives on the What, How, Why, and Who of data storytelling. We further investigated the operationalization of data stories. We identified 12 data story forms that provide concrete examples of how data stories have been presented. We derived a set of spectrum-based dimensions that capture important properties of data stories. Along each spectrum, applicable forms and design alternatives were discussed to analyze how they shape data storytelling experiences, along with data storytelling design trade-offs. Additionally, we examine how traditional narrative elements, like plot and character, have been adapted in data stories to support the operationalization of a data storytelling narratological perspective. Finally, we concluded the survey with a synthesis of our major findings and implications for future research.

cs.HC↗

arXiv2025

Fast, Slow, and Tool-augmented Thinking for LLMs: A Review

Large Language Models (LLMs) have demonstrated remarkable progress in reasoning across diverse domains. However, effective reasoning in real-world tasks requires adapting the reasoning strategy to the demands of the problem, ranging from fast, intuitive responses to deliberate, step-by-step reasoning and tool-augmented thinking. Drawing inspiration from cognitive psychology, we propose a novel taxonomy of LLM reasoning strategies along two knowledge boundaries: a fast/slow boundary separating intuitive from deliberative processes, and an internal/external boundary distinguishing reasoning grounded in the model's parameters from reasoning augmented by external tools. We systematically survey recent work on adaptive reasoning in LLMs and categorize methods based on key decision factors. We conclude by highlighting open challenges and future directions toward more adaptive, efficient, and reliable LLMs.

cs.CL↗

arXiv2025

T$^2$: An Adaptive Test-Time Scaling Strategy for Contextual Question Answering

Recent advances in Large Language Models (LLMs) have demonstrated remarkable performance in Contextual Question Answering (CQA). However, prior approaches typically employ elaborate reasoning strategies regardless of question complexity, leading to low adaptability. Recent efficient test-time scaling methods introduce budget constraints or early stop mechanisms to avoid overthinking for straightforward questions. But they add human bias to the reasoning process and fail to leverage models' inherent reasoning capabilities. To address these limitations, we present T$^2$: Think-to-Think, a novel framework that dynamically adapts reasoning depth based on question complexity. T$^2$ leverages the insight that if an LLM can effectively solve similar questions using specific reasoning strategies, it can apply the same strategy to the original question. This insight enables to adoption of concise reasoning for straightforward questions while maintaining detailed analysis for complex problems. T$^2$ works through four key steps: decomposing questions into structural elements, generating similar examples with candidate reasoning strategies, evaluating these strategies against multiple criteria, and applying the most appropriate strategy to the original question. Experimental evaluation across seven diverse CQA benchmarks demonstrates that T$^2$ not only achieves higher accuracy than baseline methods but also reduces computational overhead by up to 25.2\%.

cs.CL↗

arXiv2025

MemeReaCon: Probing Contextual Meme Understanding in Large Vision-Language Models

Memes have emerged as a popular form of multimodal online communication, where their interpretation heavily depends on the specific context in which they appear. Current approaches predominantly focus on isolated meme analysis, either for harmful content detection or standalone interpretation, overlooking a fundamental challenge: the same meme can express different intents depending on its conversational context. This oversight creates an evaluation gap: although humans intuitively recognize how context shapes meme interpretation, Large Vision Language Models (LVLMs) can hardly understand context-dependent meme intent. To address this critical limitation, we introduce MemeReaCon, a novel benchmark specifically designed to evaluate how LVLMs understand memes in their original context. We collected memes from five different Reddit communities, keeping each meme's image, the post text, and user comments together. We carefully labeled how the text and meme work together, what the poster intended, how the meme is structured, and how the community responded. Our tests with leading LVLMs show a clear weakness: models either fail to interpret critical information in the contexts, or overly focus on visual details while overlooking communicative purpose. MemeReaCon thus serves both as a diagnostic tool exposing current limitations and as a challenging benchmark to drive development toward more sophisticated LVLMs of the context-aware understanding.

cs.AI↗

arXiv2025

Stepwise Reasoning Checkpoint Analysis: A Test Time Scaling Method to Enhance LLMs' Reasoning

Mathematical reasoning through Chain-of-Thought (CoT) has emerged as a powerful capability of Large Language Models (LLMs), which can be further enhanced through Test-Time Scaling (TTS) methods like Beam Search and DVTS. However, these methods, despite improving accuracy by allocating more computational resources during inference, often suffer from path homogenization and inefficient use of intermediate results. To address these limitations, we propose Stepwise Reasoning Checkpoint Analysis (SRCA), a framework that introduces checkpoints between reasoning steps. It incorporates two key strategies: (1) Answer-Clustered Search, which groups reasoning paths by their intermediate checkpoint answers to maintain diversity while ensuring quality, and (2) Checkpoint Candidate Augmentation, which leverages all intermediate answers for final decision-making. Our approach effectively reduces path homogenization and creates a fault-tolerant mechanism by utilizing high-quality intermediate results. Experimental results show that SRCA improves reasoning accuracy compared to existing TTS methods across various mathematical datasets.

cs.CL↗

arXiv2025

ToolACE-R: Model-aware Iterative Training and Adaptive Refinement for Tool Learning

Tool learning, which allows Large Language Models (LLMs) to leverage external tools for solving complex user tasks, has emerged as a promising avenue for extending model capabilities. However, existing approaches primarily focus on data synthesis for fine-tuning LLMs to invoke tools effectively, largely ignoring how to fully stimulate the potential of the model. In this paper, we propose ToolACE-R, a novel framework that includes both model-aware iterative training and adaptive refinement for tool learning. ToolACE-R features a model-aware iterative training procedure that progressively adjust training samples based on the model's evolving capabilities to maximize its potential. Additionally, it incorporates self-refinement training corpus which emphasizes LLM's ability to iteratively refine their tool calls, optimizing performance without requiring external feedback. Furthermore, we introduce adaptive self-refinement mechanism for efficient test-time scaling, where the trained model can autonomously determine when to stop the process based on iterative self-refinement. We conduct extensive experiments across several benchmark datasets, showing that ToolACE-R achieves competitive performance compared to advanced API-based models. The performance of tool invocation can be further improved efficiently through adaptive self-refinement. These results highlight the effectiveness and generalizability of ToolACE-R, offering a promising direction for more efficient and scalable tool learning.

cs.CL↗

arXiv2025

FReM: A Flexible Reasoning Mechanism for Balancing Quick and Slow Thinking in Long-Context Question Answering

Long-context question-answering (LCQA) systems have greatly benefited from the powerful reasoning capabilities of large language models (LLMs), which can be categorized into slow and quick reasoning modes. However, both modes have their limitations. Slow thinking generally leans to explore every possible reasoning path, which leads to heavy overthinking and wastes time. Quick thinking usually relies on pattern matching rather than truly understanding the query logic, which misses proper understanding. To address these issues, we propose FReM: Flexible Reasoning Mechanism, a method that adjusts reasoning depth according to the complexity of each question. Specifically, FReM leverages synthetic reference QA examples to provide an explicit chain of thought, enabling efficient handling of simple queries while allowing deeper reasoning for more complex ones. By doing so, FReM helps quick-thinking models move beyond superficial pattern matching and narrows the reasoning space for slow-thinking models to avoid unnecessary exploration. Experiments on seven QA datasets show that FReM improves reasoning accuracy and scalability, particularly for complex multihop questions, indicating its potential to advance LCQA methodologies.

cs.CL↗

arXiv2024

ToolFlow: Boosting LLM Tool-Calling Through Natural and Coherent Dialogue Synthesis

Supervised fine-tuning (SFT) is a common method to enhance the tool calling capabilities of Large Language Models (LLMs), with the training data often being synthesized. The current data synthesis process generally involves sampling a set of tools, formulating a requirement based on these tools, and generating the call statements. However, tools sampled randomly lack relevance, making them difficult to combine and thus reducing the diversity of the data. Additionally, current work overlooks the coherence between turns of dialogues, leading to a gap between the synthesized data and real-world scenarios. To address these issues, we propose a Graph-based Sampling strategy to sample more relevant tool combinations, and a Planned-generation strategy to create plans that guide the synthesis of coherent dialogues. We integrate these two strategies and enable multiple agents to synthesize the dialogue data interactively, resulting in our tool-calling data synthesis pipeline ToolFlow. Data quality assessments demonstrate improvements in the naturalness and coherence of our synthesized dialogues. Finally, we apply SFT on LLaMA-3.1-8B using 8,000 synthetic dialogues generated with ToolFlow. Results show that the model achieves tool-calling performance comparable to or even surpassing GPT-4, while maintaining strong general capabilities.

cs.CL↗

arXiv2024

MlingConf: A Comprehensive Study of Multilingual Confidence Estimation on Large Language Models

The tendency of Large Language Models (LLMs) to generate hallucinations raises concerns regarding their reliability. Therefore, confidence estimations indicating the extent of trustworthiness of the generations become essential. However, current LLM confidence estimations in languages other than English remain underexplored. This paper addresses this gap by introducing a comprehensive investigation of Multilingual Confidence estimation (MlingConf) on LLMs, focusing on both language-agnostic (LA) and language-specific (LS) tasks to explore the performance and language dominance effects of multilingual confidence estimations on different tasks. The benchmark comprises four meticulously checked and human-evaluate high-quality multilingual datasets for LA tasks and one for the LS task tailored to specific social, cultural, and geographical contexts of a language. Our experiments reveal that on LA tasks English exhibits notable linguistic dominance in confidence estimations than other languages, while on LS tasks, using question-related language to prompt LLMs demonstrates better linguistic dominance in multilingual confidence estimations. The phenomena inspire a simple yet effective native-tone prompting strategy by employing language-specific prompts for LS tasks, effectively improving LLMs' reliability and accuracy on LS tasks.

cs.CL↗

arXiv2024

Social Network Datasets on Reddit Financial Discussion

Stock markets are impacted by a large variety of factors including news and discussions among investors about investment opportunities. With the emergence of social media, new opportunities for having financial discussions arose. The market frenzy surrounding GameStop (GME) on the Reddit subreddit Wallstreetbets, caused financial discussion forums to receive widespread attention and it was established that Wallstreetbets played a leading role in the stock market movements of GME. Here, we present a new data set for exploring the effect of social media discussion forums on the stock market. The dataset consists of posts published on various Reddit subreddits concerning the popular meme stocks GameStop (GME), American Multi-Cinema Entertainment Holdings (AMC), and BlackBerry (BB). We document the data collection and processing steps and show that the posts and comments about these meme stocks are related to their market movements.

cs.SI↗