Yuanzhe Shen

Yixu Huang, Bo Li, Na Li et al.

Generating a game is not the same as making one that can be played. Despite advances in code generation, existing approaches treat game generation as one-shot translation from prompt to artifact, leaving interaction-level failures undetected. We argue that evaluating and improving game generation requires a player, and study two roles for graphical user interface (GUI) agents in this process: (1) as an objective evaluator, for which we introduce PlaytestArena, a new evaluation environment that pairs 200 browser-based game generation tasks across eight genres with rubrics of expected in-play behaviors, adjudicated by a GUI agent that loads each build in a browser and plays it; and (2) as a subjective playtester, for which we propose Play2Code, where a game agent and a GUI agent operate in a sustained loop with shared memory, turning game generation into a dialogue between coding and playing. Our experiments show that even frontier models struggle to generate playable games directly, while Play2Code achieves a 66.8\% rubric pass-rate, improving over single-pass and agentic-coding baselines by 37.1 and 14.6 points respectively. Further analysis shows that GUI playtester feedback is more traceable than a human report, yet idiosyncratic in ways reminiscent of human testers, establishing game playtesting as a critical testbed for interactive code generation. Our project website is available at https://continual-game-generation.vercel.app/.

Yuanzhe Shen, Kaimin Wang, Changze Lv et al.

The continuous evolution and enhanced reasoning capabilities of large language models (LLMs) have elevated their role in complex tasks, notably in travel planning, where demand for personalized, high-quality itineraries is rising. However, current benchmarks often rely on unrealistic simulated data, failing to reflect the differences between LLM-generated and real-world itineraries. Existing evaluation metrics, which primarily emphasize constraints, fall short of providing a comprehensive assessment of the overall quality of travel plans. To address these limitations, we introduce TripTailor, a benchmark designed specifically for personalized travel planning in real-world scenarios. This dataset features an extensive collection of over 500,000 real-world points of interest (POIs) and nearly 4,000 diverse travel itineraries, complete with detailed information, providing a more authentic evaluation framework. Experiments show that fewer than 10\% of the itineraries generated by the latest state-of-the-art LLMs achieve human-level performance. Moreover, we identify several critical challenges in travel planning, including the feasibility, rationality, and personalized customization of the proposed solutions. We hope that TripTailor will drive the development of travel planning agents capable of understanding and meeting user needs while generating practical itineraries. Our code and dataset are available at https://github.com/swxkfm/TripTailor

Muzhao Tian, Zisu Huang, Xiaohua Wang et al.

As LLM-based agents are increasingly used in long-term interactions, cumulative memory is critical for enabling personalization and maintaining stylistic consistency. However, most existing systems adopt an ``all-or-nothing'' approach to memory usage: incorporating all relevant past information can lead to \textit{Memory Anchoring}, where the agent is trapped by past interactions, while excluding memory entirely results in under-utilization and the loss of important interaction history. We show that an agent's reliance on memory can be modeled as an explicit and user-controllable dimension. We first introduce a behavioral metric of memory dependence to quantify the influence of past interactions on current outputs. We then propose \textbf{Stee}rable \textbf{M}emory Agent, \texttt{SteeM}, a framework that allows users to dynamically regulate memory reliance, ranging from a fresh-start mode that promotes innovation to a high-fidelity mode that closely follows interaction history. Experiments across different scenarios demonstrate that our approach consistently outperforms conventional prompting and rigid memory masking strategies, yielding a more nuanced and effective control for personalized human-agent collaboration.

Yuanzhe Shen, Zisu Huang, Zhengyuan Wang et al.

As LLM-based agents are deployed in increasingly complex real-world settings, existing benchmarks underrepresent key challenges such as enforcing global constraints, coordinating multi-tool reasoning, and adapting to evolving user behavior over long, multi-turn interactions. To bridge this gap, we introduce \textbf{TRIP-Bench}, a long-horizon benchmark grounded in realistic travel-planning scenarios. TRIP-Bench leverages real-world data, offers 18 curated tools and 40+ travel requirements, and supports automated evaluation. It includes splits of varying difficulty; the hard split emphasizes long and ambiguous interactions, style shifts, feasibility changes, and iterative version revision. Dialogues span up to 15 user turns, can involve 150+ tool calls, and may exceed 200k tokens of context. Experiments show that even advanced models achieve at most 50\% success on the easy split, with performance dropping below 10\% on hard subsets. We further propose \textbf{GTPO}, an online multi-turn reinforcement learning method with specialized reward normalization and reward differencing. Applied to Qwen2.5-32B-Instruct, GTPO improves constraint satisfaction and interaction robustness, outperforming Gemini-3-Pro in our evaluation. We expect TRIP-Bench to advance practical long-horizon interactive agents, and GTPO to provide an effective online RL recipe for robust long-horizon training.

Yuanzhe Shen, Zisu Huang, Zhengkang Guo et al.

The rapid advancement of large language models (LLMs) has driven their adoption across diverse domains, yet their ability to generate harmful content poses significant safety challenges. While extensive research has focused on mitigating harmful outputs, such efforts often come at the cost of excessively rejecting harmless prompts. Striking a balance among safety, over-refusal, and utility remains a critical challenge. In this work, we introduce IntentionReasoner, a novel safeguard mechanism that leverages a dedicated guard model to perform intent reasoning, multi-level safety classification, and query rewriting to neutralize potentially harmful intent in edge-case queries. Specifically, we first construct a comprehensive dataset comprising approximately 163,000 queries, each annotated with intent reasoning, safety labels, and rewritten versions. Supervised fine-tuning is then applied to equip the guard model with foundational capabilities in format adherence, intent analysis, and safe rewriting. Finally, we apply a tailored multi-reward optimization strategy that integrates rule-based heuristics and reward model signals within a reinforcement learning framework to further enhance performance. Extensive experiments show that IntentionReasoner excels in multiple safeguard benchmarks, generation quality evaluations, and jailbreak attack scenarios, significantly enhancing safety while effectively reducing over-refusal rates and improving the quality of responses.

Zhengkang Guo, Wenhao Liu, Mingchen Xie et al.

Large language models (LLMs) are increasingly expected to tackle complex tasks, driven by their expanding applications and users' growing proficiency in crafting sophisticated prompts. However, as the number of explicitly stated requirements increases (particularly more than 10 constraints), LLMs often struggle to accurately follow such complex instructions, which limits their applicability in complex real-world scenarios. To the best of our knowledge, existing datasets do not exceed 10 constraints per instance. To address this challenge, we propose RECAST, an efficient and scalable framework for synthesizing datasets where each example incorporates far more constraints than those in existing benchmarks, aiming to challenge and extend the boundaries of models' ability to follow complex instructions. These constraints are extracted from real-world prompt-response pairs to ensure practical relevance. Using this framework, we construct RECAST-30K, a large-scale, high-quality dataset comprising 30k instances spanning 19 constraint types. Experimental results demonstrate that models finetuned on RECAST-30K substantially improve in following complex instructions while maintaining their general capabilities without degradation. Moreover, RECAST enables automatic verification of constraint satisfaction via rule-based validators for quantitative constraints and LLM-based validators for qualitative ones; the verifiability provided by RECAST enables the design of reward functions for reinforcement learning, which further boosts model performance on complex and challenging tasks.

Yuanzhe Shen, Yide Liu, Zisu Huang et al.

Large language models (LLMs) demonstrate remarkable performance across diverse tasks, yet their effectiveness frequently depends on costly commercial APIs or cloud services. Model selection thus entails a critical trade-off between performance and cost: high-performing LLMs typically incur substantial expenses, whereas budget-friendly small language models (SLMs) are constrained by limited capabilities. Current research primarily proposes two routing strategies: pre-generation routing and cascade routing. Both approaches have distinct characteristics, with cascade routing typically offering superior cost-effectiveness and accuracy despite its higher latency. To further address the limitations of both approaches, we introduce SATER, a dual-mode compatible approach that fine-tunes models through shortest-response preference optimization and a confidence-aware rejection mechanism. SATER significantly reduces redundant outputs and response times, while improving both the performance of pre-generation routing and the efficiency of cascade routing. Experiments across three SLMs and six datasets, varying in type and complexity, demonstrate that SATER achieves comparable performance while consistently reducing computational costs by over 50\% and cascade latency by over 80\%.