failure-aware planning

This work addresses the underutilization of failed experiments as reusable knowledge assets in AI-assisted scientific research. It proposes a “negative knowledge memory layer” that employs curation agents to structure and store unsuccessful experimental attempts in a shared repository, enabling downstream research agents to explicitly accept or reject these records when designing new experiments. For the first time, structured negative knowledge is treated as an independent knowledge asset, explicitly maintained and transferred across tasks. Integrated into the AutoResearch multi-agent framework, the approach supports the generation, typologized storage, and retrieval of negative knowledge. Evaluated on ScienceAgentBench and nonlinear mathematical physics PDE tasks, the method solves novel problems that baseline systems cannot address—using fewer tokens—and substantially improves cross-problem exploration efficiency.

Recommender systems face persistent challenges including filter bubbles, sparse user-item interactions, cold-start problems, and feedback loops. Existing approaches predominantly leverage positive behavioral signals while underutilizing the critical role of negative feedback in preference modeling. This survey establishes, for the first time, a taxonomy of negative sampling methodologies—categorizing over one hundred works into five paradigmatic classes. It further introduces a scenario-aware adaptation framework that elucidates the pivotal roles of negative sampling in modeling dynamic preferences, mitigating filter bubbles, and alleviating feedback bias. Finally, it identifies three emerging research frontiers: enhancing interpretability, integrating causal inference, and synergizing with large language models. By rigorously delineating the theoretical boundaries and practical implementation pathways of negative sampling, this work elevates it from an empirical engineering heuristic to a foundational paradigm in recommender system design.

This study challenges the foundational assumption in mainstream machine learning that objective ground-truth labels exist, an assumption often violated in real-world scenarios and leading to inaccurate evaluation and learning. Adopting a negative ontological stance—asserting that no single true label exists—the work introduces, for the first time, this philosophical perspective into machine learning through a democratic supervision framework. It proposes representing each instance with multiple imprecise yet authentic truth labels (MIATTs), accompanied by a logic-driven mechanism for generating and evaluating MIATTs, as well as a truth-learning strategy that operates without relying on a predefined ground truth. The resulting EL-MIATTs methodology is validated in real educational settings, demonstrating not only a departure from the conventional single-ground-truth paradigm but also practical utility in supporting personalized education and career development.

To address suboptimal decision-making in real-world project selection arising from human cognitive limitations, this paper proposes a teachable strategy discovery method explicitly designed for human cognitive constraints and develops an interactive intelligent tutoring system to enhance practical decision-making competence. Methodologically, it pioneers the application of automated strategy discovery to authentic project selection tasks, introducing the MGPS (Model-Guided Policy Search) algorithm and an interpretable, pedagogically grounded strategy generation framework that integrates cognitive-model-informed policy optimization with computationally rigorous benchmark evaluation. Results demonstrate that MGPS consistently outperforms state-of-the-art methods in both solution quality and computational efficiency. Moreover, human participants trained via the intelligent tutor exhibit statistically significant improvements in strategy quality, empirically validating the method’s effectiveness and practical utility in improving human decision-making under naturalistic conditions.

This study addresses the challenge of translating maladaptive cognitive cycles—such as rumination and repetitive regret—into constructive behavioral change. Integrating the Transtheoretical Model with Gross’s process model of emotion regulation, this work operationalizes the latter for the first time to design and implement a voice diary system. The system synthesizes counterfactual thinking, “if-then” implementation intentions (WhatIf-Planning), and structured reflective prompts into a unified reflection-to-action framework, effectively facilitating users’ progression from the preparation stage to actual behavior change. Experimental results demonstrate that this approach significantly enhances coping flexibility: participants in the Gross-guided condition generated a greater number of counterfactual alternatives, formulated more specific and actionable plans, and engaged more proactively in self-initiated behavioral adjustments.

This study addresses the challenge that individuals in social networks may emulate negative role models due to uncertainty about their true labels, thereby undermining social welfare. To mitigate this, the authors investigate how a social planner can strategically disclose role model labels under a limited disclosure budget to steer users toward positive exemplars and maximize overall welfare. They propose a proxy welfare function that preserves submodularity—overcoming the issue that revealing negative labels typically breaks submodularity—and design a constant-factor approximation algorithm. Novel mechanisms are introduced to ensure fairness, prioritize intervention for high-risk individuals, and expand coverage radius. Experiments on four real-world datasets demonstrate that the approach achieves a constant approximation ratio even in the presence of a constant number of negative neighbors, while guaranteeing equitable welfare gains across diverse population groups.

Current understanding of how reinforcement learning (RL) enhances reasoning capabilities during post-training remains unclear. This study addresses this gap through controlled mathematical reasoning experiments, explicitly disentangling and validating two core mechanisms in post-training: policy selection and policy improvement. Leveraging the Qwen-2.5-1.5B model with diverse supervised fine-tuning (SFT) data and progressively harder RL data, the research demonstrates that diverse SFT data effectively facilitates policy selection, while high-difficulty RL data drives policy improvement. The findings not only clarify the distinct roles of SFT and RL data in activating these mechanisms but also offer actionable pathways for enhancing model reasoning performance.

This study addresses the tendency of users to uncritically accept recommendations from AI decision-support systems, which can lead to erroneous judgments. To mitigate this issue, the authors propose a data-driven prompting mechanism that enhances prospective reasoning in human-AI collaboration by automatically generating reflective questions via large language models (LLMs). The approach integrates a structured question taxonomy, a cognitive engagement scale, and LLM-based question generation. A prototype system was developed and evaluated in a clinical setting, combining methods from LLMs, human-computer interaction design, and cognitive assessment. Empirical results demonstrate that the proposed mechanism significantly improves clinicians’ critical appraisal of AI-generated outputs, eliciting positive user feedback and offering a novel pathway toward developing AI systems that function as “thinking tools” rather than mere decision aids.