Quan Shi

Haocheng Luo, Jiahui Liu, Ruicheng Zhang et al.

While vision-language models excel at general multimodal understanding, they still struggle with visual spatial planning. We attribute this to a perception-reasoning modality gap: visual planning requires models to infer latent state structures from pixels and then reason over the recovered structure to produce valid actions, whereas symbolic planning directly leverages explicit objects and constraints. This creates dual bottlenecks in visual state recovery and multi-step planning. To address this, we propose MGSD, a two-stage modality-gap-aware self-distillation framework. First, a cold-start grounding stage equips the visual student with reliable state representations, minimizing early perception noise. Second, a privileged teacher transfers planning capabilities via on-policy distillation, using explicit symbolic states to supervise the student's own visual rollout prefixes. Crucially, symbolic data is used strictly during training, leaving inference purely visual. Experiments on visual planning benchmarks show that MGSD consistently improves visual planning across both 4B and 8B backbones, raising the macro average by 19.3% and 18.4%, respectively. The resulting models narrow the gap to symbolic-input upper bounds, while ablations and diagnostics confirm that the improvement comes from both visual state recovery and optimal-path reasoning. These results suggest that modality-gap-aware self-distillation improves not only how models perceive actionable states, but also how they plan over the inferred structure. Code is available at https://github.com/Oranger-l/MGSD.

Hongjin Su, Howard Yen, Mengzhou Xia et al.

Existing retrieval benchmarks primarily consist of information-seeking queries (e.g., aggregated questions from search engines) where keyword or semantic-based retrieval is usually sufficient. However, many complex real-world queries require in-depth reasoning to identify relevant documents that go beyond surface form matching. For example, finding documentation for a coding question requires understanding the logic and syntax of the functions involved. To better benchmark retrieval on such challenging queries, we introduce BRIGHT, the first text retrieval benchmark that requires intensive reasoning to retrieve relevant documents. Our dataset consists of 1,384 real-world queries spanning diverse domains, such as economics, psychology, mathematics, and coding. These queries are drawn from naturally occurring and carefully curated human data. Extensive evaluation reveals that even state-of-the-art retrieval models perform poorly on BRIGHT. The leading model on the MTEB leaderboard (Muennighoff et al., 2023) SFR-Embedding-Mistral (Meng et al., 2024), which achieves a score of 59.0 nDCG@10,1 produces a score of nDCG@10 of 18.3 on BRIGHT. We show that incorporating explicit reasoning about the query improves retrieval performance by up to 12.2 points. Moreover, incorporating retrieved documents from the top-performing retriever boosts question-answering performance. We believe that BRIGHT paves the way for future research on retrieval systems in more realistic and challenging settings.

Fengwei Teng, Zhaoyang Yu, Quan Shi et al.

Large Language Models (LLMs) achieve superior performance through training-time scaling, and test-time scaling further enhances their capabilities by conducting effective reasoning during inference. However, as the scale of reasoning increases, existing test-time scaling methods suffer from accumulated historical information, which not only wastes computational resources but also interferes with effective reasoning. To address this issue, we observe that complex reasoning can be achieved by solving a series of independent and self-contained subquestions. These subquestions are essentially \textit{atomic questions}, exhibiting the memoryless property similar to Markov processes. Based on this observation, we propose Atom of Thoughts (\our), where each state transition consists of decomposing the current question into a dependency-based directed acyclic graph and contracting its subquestions, forming a simplified question that maintains answer equivalence with the original problem. This answer preservation enables the iterative \textit{decomposition-contraction} process to naturally form a meaningful Markov reasoning process. Furthermore, these atomic states can be seamlessly integrated into existing test-time scaling methods, enabling \our to serve as a plug-in enhancement for improving reasoning capabilities. Experiments across six benchmarks demonstrate the effectiveness of \our both as a standalone framework and a plug-in enhancement. Notably, on HotpotQA, when applied to gpt-4o-mini, \our achieves an \textbf{80.6\%} F1 score, surpassing o3-mini by \textbf{3.4\%} and DeepSeek-R1 by \textbf{10.6\%}. The code is available at \href{https://github.com/qixucen/atom}{https://github.com/qixucen/atom}.

Quan Shi, Alexandra Zytek, Pedram Razavi et al.

Conversational agents are increasingly deployed in knowledge-intensive settings, where correct behavior depends on retrieving and applying domain-specific knowledge from large, proprietary, and unstructured corpora during live interactions with users. Yet most existing benchmarks evaluate retrieval or tool use independently of each other, creating a gap in realistic, fully agentic evaluation over unstructured data in long-horizon interactions. We introduce $τ$-Knowledge, an extension of $τ$-Bench for evaluating agents in environments where success depends on coordinating external, natural-language knowledge with tool outputs to produce verifiable, policy-compliant state changes. Our new domain, $τ$-Banking, models realistic fintech customer support workflows in which agents must navigate roughly 700 interconnected knowledge documents while executing tool-mediated account updates. Across embedding-based retrieval and terminal-based search, even frontier models with high reasoning budgets achieve only $\sim$25.5% pass^1, with reliability degrading sharply over repeated trials. Agents struggle to retrieve the correct documents from densely interlinked knowledge bases and to reason accurately over complex internal policies. Overall, $τ$-Knowledge provides a realistic testbed for developing agents that integrate unstructured knowledge in human-facing deployments.

Quan Shi, Carlos E. Jimenez, Shunyu Yao et al. · princeton

Recent advancements in AI reasoning have driven substantial improvements across diverse tasks. A critical open question is whether these improvements also yields better knowledge transfer: the ability of models to communicate reasoning in ways humans can understand, apply, and learn from. To investigate this, we introduce Knowledge Integration and Transfer Evaluation (KITE), a conceptual and experimental framework for Human-AI knowledge transfer capabilities and conduct the first large-scale human study (N=118) explicitly designed to measure it. In our two-phase setup, humans first ideate with an AI on problem-solving strategies, then independently implement solutions, isolating model explanations' influence on human understanding. Our findings reveal that although model benchmark performance correlates with collaborative outcomes, this relationship is notably inconsistent, featuring significant outliers, indicating that knowledge transfer requires dedicated optimization. Our analysis identifies behavioral and strategic factors mediating successful knowledge transfer. We release our code, dataset, and evaluation framework to support future work on communicatively aligned models.

Quan Shi, Carlos E. Jimenez, Stephen Dong et al. · princeton

As language models achieve increasingly human-like capabilities in conversational text generation, a critical question emerges: to what extent can these systems simulate the characteristics of specific individuals? To evaluate this, we introduce IMPersona, a framework for evaluating LMs at impersonating specific individuals' writing style and personal knowledge. Using supervised fine-tuning and a hierarchical memory-inspired retrieval system, we demonstrate that even modestly sized open-source models, such as Llama-3.1-8B-Instruct, can achieve impersonation abilities at concerning levels. In blind conversation experiments, participants (mis)identified our fine-tuned models with memory integration as human in 44.44% of interactions, compared to just 25.00% for the best prompting-based approach. We analyze these results to propose detection methods and defense strategies against such impersonation attempts. Our findings raise important questions about both the potential applications and risks of personalized language models, particularly regarding privacy, security, and the ethical deployment of such technologies in real-world contexts.

Runyu Wang, Peng Ping, Zhengyu Guo et al.

Fine-tuning adapts pretrained models for specific tasks but poses the risk of catastrophic forgetting (CF), where critical knowledge from pre-training is overwritten. Current Parameter-Efficient Fine-Tuning (PEFT) methods for Large Language Models (LLMs), while efficient, often sacrifice general capabilities. To address the issue of CF in a general-purpose PEFT framework, we propose \textbf{Lo}w-damage \textbf{K}nowledge \textbf{I}mplanting (\textbf{LoKI}), a PEFT technique that is based on a mechanistic understanding of how knowledge is stored in transformer architectures. In two real-world scenarios, LoKI demonstrates task-specific performance that is comparable to or even surpasses that of full fine-tuning and LoRA-based methods across various model types, while significantly better preserving general capabilities. Our work connects mechanistic insights into LLM knowledge storage with practical fine-tuning objectives, achieving state-of-the-art trade-offs between task specialization and the preservation of general capabilities. Our implementation is publicly available as ready-to-use code\footnote{https://github.com/Nexround/LoKI}.

Quan Shi, Michael Tang, Karthik Narasimhan et al.

Computing olympiads contain some of the most challenging problems for humans, requiring complex algorithmic reasoning, puzzle solving, in addition to generating efficient code. However, it has been understudied as a domain to evaluate language models (LMs). In this paper, we introduce the USACO benchmark with 307 problems from the USA Computing Olympiad, along with high-quality unit tests, reference code, and official analyses for each problem. These resources enable us to construct and test a range of LM inference methods for competitive programming for the first time. We find GPT-4 only achieves a 8.7% pass@1 accuracy with zero-shot chain-of-thought prompting, and our best inference method improves it to 20.2% using a combination of self-reflection and retrieval over episodic knowledge. However, this is far from solving the benchmark. To better understand the remaining challenges, we design a novel human-in-the-loop study and surprisingly find that a small number of targeted hints enable GPT-4 to solve 13 out of 15 problems previously unsolvable by any model and method. Our benchmark, baseline methods, quantitative results, and qualitative analysis serve as an initial step toward LMs with grounded, creative, and algorithmic reasoning.

Yanjie Sun, Zhike Gong, Quan Shi et al.

Effective control of credit risk is a key link in the steady operation of commercial banks. This paper is mainly based on the customer information dataset of a foreign commercial bank in Kaggle, and we use LightGBM algorithm to build a classifier to classify customers, to help the bank judge the possibility of customer credit default. This paper mainly deals with characteristic engineering, such as missing value processing, coding, imbalanced samples, etc., which greatly improves the machine learning effect. The main innovation of this paper is to construct new feature attributes on the basis of the original dataset so that the accuracy of the classifier reaches 0.734, and the AUC reaches 0.772, which is more than many classifiers based on the same dataset. The model can provide some reference for commercial banks' credit granting, and also provide some feature processing ideas for other similar studies.

Quan Shi, Wang Xi, Zenghui Ding et al.

The illusion phenomenon of large language models (LLMs) is the core obstacle to their reliable deployment. This article formalizes the large language model as a probabilistic Turing machine by constructing a "computational necessity hierarchy", and for the first time proves the illusions are inevitable on diagonalization, incomputability, and information theory boundaries supported by the new "learner pump lemma". However, we propose two "escape routes": one is to model Retrieval Enhanced Generations (RAGs) as oracle machines, proving their absolute escape through "computational jumps", providing the first formal theory for the effectiveness of RAGs; The second is to formalize continuous learning as an "internalized oracle" mechanism and implement this path through a novel neural game theory framework.Finally, this article proposes a

Wang Xi, Quan Shi, Tian Yu et al.

Automated generation of high-quality media presentations is challenging, requiring robust content extraction, narrative planning, visual design, and overall quality optimization. Existing methods often produce presentations with logical inconsistencies and suboptimal layouts, thereby struggling to meet professional standards. To address these challenges, we introduce RCPS (Reflective Coherent Presentation Synthesis), a novel framework integrating three key components: (1) Deep Structured Narrative Planning; (2) Adaptive Layout Generation; (3) an Iterative Optimization Loop. Additionally, we propose PREVAL, a preference-based evaluation framework employing rationale-enhanced multi-dimensional models to assess presentation quality across Content, Coherence, and Design. Experimental results demonstrate that RCPS significantly outperforms baseline methods across all quality dimensions, producing presentations that closely approximate human expert standards. PREVAL shows strong correlation with human judgments, validating it as a reliable automated tool for assessing presentation quality.