Grounded Chess Reasoning in Language Models via Master Distillation

Language models often lack grounded reasoning capabilities in specialized domains where training data is scarce but bespoke systems excel. We introduce a general framework for distilling expert system reasoning into natural language chain-of-thought explanations, enabling compact models to acquire domain expertise and the ability to generate faithful, grounded explanations. Rather than distilling only final outputs, we capture the full reasoning process, transforming opaque expert computations into transparent, step-by-step explanations. We demonstrate this approach in chess, a canonical reasoning domain where language models continue to underperform. Our 4B parameter model, C1, advances from a near-zero baseline to 48.1% accuracy, outperforming all open-source models and most frontier proprietary systems. Notably, C1 surpasses its distillation teacher and generates solutions in two orders of magnitude fewer tokens than baselines. Unlike prior neural chess approaches that predict only best moves, C1 generates explainable solutions revealing strategic reasoning. Our pipeline combines supervised fine-tuning and reinforcement learning with theme-balanced data sampling for comprehensive tactical coverage. Master Distillation demonstrates how to inject expert-level knowledge into compact models for under-optimized domains, offering a recipe for unlocking RLVR where LLMs lack sufficient base capabilities.

Level Up: Defining and Exploiting Transitional Problems for Curriculum Learning

Curriculum learning--ordering training examples in a sequence to aid machine learning--takes inspiration from human learning, but has not gained widespread acceptance. Static strategies for scoring item difficulty rely on indirect proxy scores of varying quality and produce curricula that are not specific to the learner at hand. Dynamic approaches base difficulty estimates on gradient information, requiring considerable extra computation during training. We introduce a novel method for measuring the difficulty of individual problem instances directly relative to the ability of a given model, and identify transitional problems that are consistently easier as model ability increases. Applying this method to chess and mathematics, we find that training on a curriculum that "levels up" from easier to harder transitional problems most efficiently improves a model to the next tier of competence. These problems induce a natural progression from easier to harder items, which outperforms other training strategies. By measuring difficulty directly relative to model competence, our method yields interpretable problems, learner-specific curricula, and a principled basis for step-by-step improvement.

SEAM: Semantically Equivalent Across Modalities Benchmark for Vision-Language Models

Evaluating whether vision-language models (VLMs) reason consistently across representations is challenging because modality comparisons are typically confounded by task differences and asymmetric information. We introduce SEAM, a benchmark that pairs semantically equivalent inputs across four domains that have existing standardized textual and visual notations. By employing distinct notation systems across modalities, in contrast to OCR-based image-text pairing, SEAM provides a rigorous comparative assessment of the textual-symbolic and visual-spatial reasoning capabilities of VLMs. Across 21 contemporary models, we observe systematic modality imbalance: vision frequently lags language in overall performance, despite the problems containing semantically equivalent information, and cross-modal agreement is relatively low. Our error analysis reveals two main drivers: textual perception failures from tokenization in domain notation and visual perception failures that induce hallucinations. We also show that our results are largely robust to visual transformations. SEAM establishes a controlled, semantically equivalent setting for measuring and improving modality-agnostic reasoning.

Maia-2: A Unified Model for Human-AI Alignment in Chess

There are an increasing number of domains in which artificial intelligence (AI) systems both surpass human ability and accurately model human behavior. This introduces the possibility of algorithmically-informed teaching in these domains through more relatable AI partners and deeper insights into human decision-making. Critical to achieving this goal, however, is coherently modeling human behavior at various skill levels. Chess is an ideal model system for conducting research into this kind of human-AI alignment, with its rich history as a pivotal testbed for AI research, mature superhuman AI systems like AlphaZero, and precise measurements of skill via chess rating systems. Previous work in modeling human decision-making in chess uses completely independent models to capture human style at different skill levels, meaning they lack coherence in their ability to adapt to the full spectrum of human improvement and are ultimately limited in their effectiveness as AI partners and teaching tools. In this work, we propose a unified modeling approach for human-AI alignment in chess that coherently captures human style across different skill levels and directly captures how people improve. Recognizing the complex, non-linear nature of human learning, we introduce a skill-aware attention mechanism to dynamically integrate players' strengths with encoded chess positions, enabling our model to be sensitive to evolving player skill. Our experimental results demonstrate that this unified framework significantly enhances the alignment between AI and human players across a diverse range of expertise levels, paving the way for deeper insights into human decision-making and AI-guided teaching tools.

DiffuDetox: A Mixed Diffusion Model for Text Detoxification

Text detoxification is a conditional text generation task aiming to remove offensive content from toxic text. It is highly useful for online forums and social media, where offensive content is frequently encountered. Intuitively, there are diverse ways to detoxify sentences while preserving their meanings, and we can select from detoxified sentences before displaying text to users. Conditional diffusion models are particularly suitable for this task given their demonstrated higher generative diversity than existing conditional text generation models based on language models. Nonetheless, text fluency declines when they are trained with insufficient data, which is the case for this task. In this work, we propose DiffuDetox, a mixed conditional and unconditional diffusion model for text detoxification. The conditional model takes toxic text as the condition and reduces its toxicity, yielding a diverse set of detoxified sentences. The unconditional model is trained to recover the input text, which allows the introduction of additional fluent text for training and thus ensures text fluency. Extensive experimental results and in-depth analysis demonstrate the effectiveness of our proposed DiffuDetox.

Bayesian Knowledge-driven Critiquing with Indirect Evidence

Conversational recommender systems (CRS) enhance the expressivity and personalization of recommendations through multiple turns of user-system interaction. Critiquing is a well-known paradigm for CRS that allows users to iteratively refine recommendations by providing feedback about attributes of recommended items. While existing critiquing methodologies utilize direct attributes of items to address user requests such as 'I prefer Western movies', the opportunity of incorporating richer contextual and side information about items stored in Knowledge Graphs (KG) into the critiquing paradigm has been overlooked. Employing this substantial knowledge together with a well-established reasoning methodology paves the way for critique-based recommenders to allow for complex knowledge-based feedback (e.g., 'I like movies featuring war side effects on veterans') which may arise in natural user-system conversations. In this work, we aim to increase the flexibility of critique-based recommendation by integrating KGs and propose a novel Bayesian inference framework that enables reasoning with relational knowledge-based feedback. We study and formulate the framework considering a Gaussian likelihood and evaluate it on two well-known recommendation datasets with KGs. Our evaluations demonstrate the effectiveness of our framework in leveraging indirect KG-based feedback (i.e., preferred relational properties of items rather than preferred items themselves), often improving personalized recommendations over a one-shot recommender by more than 15%. This work enables a new paradigm for using rich knowledge content and reasoning over indirect evidence as a mechanism for critiquing interactions with CRS.

LogicRec: Recommendation with Users' Logical Requirements

Users may demand recommendations with highly personalized requirements involving logical operations, e.g., the intersection of two requirements, where such requirements naturally form structured logical queries on knowledge graphs (KGs). To date, existing recommender systems lack the capability to tackle users' complex logical requirements. In this work, we formulate the problem of recommendation with users' logical requirements (LogicRec) and construct benchmark datasets for LogicRec. Furthermore, we propose an initial solution for LogicRec based on logical requirement retrieval and user preference retrieval, where we face two challenges. First, KGs are incomplete in nature. Therefore, there are always missing true facts, which entails that the answers to logical requirements can not be completely found in KGs. In this case, item selection based on the answers to logical queries is not applicable. We thus resort to logical query embedding (LQE) to jointly infer missing facts and retrieve items based on logical requirements. Second, answer sets are under-exploited. Existing LQE methods can only deal with query-answer pairs, where queries in our case are the intersected user preferences and logical requirements. However, the logical requirements and user preferences have different answer sets, offering us richer knowledge about the requirements and preferences by providing requirement-item and preference-item pairs. Thus, we design a multi-task knowledge-sharing mechanism to exploit these answer sets collectively. Extensive experimental results demonstrate the significance of the LogicRec task and the effectiveness of our proposed method.

FALCON: Sound and Complete Neural Semantic Entailment over ALC Ontologies

Many ontologies, i.e., Description Logic (DL) knowledge bases, have been developed to provide rich knowledge about various domains, and a lot of them are based on ALC, i.e., a prototypical and expressive DL, or its extensions. The main task that explores ALC ontologies is to compute semantic entailment. Symbolic approaches can guarantee sound and complete semantic entailment but are sensitive to inconsistency and missing information. To this end, we propose FALCON, a Fuzzy ALC Ontology Neural reasoner. FALCON uses fuzzy logic operators to generate single model structures for arbitrary ALC ontologies, and uses multiple model structures to compute semantic entailments. Theoretical results demonstrate that FALCON is guaranteed to be a sound and complete algorithm for computing semantic entailments over ALC ontologies. Experimental results show that FALCON enables not only approximate reasoning (reasoning over incomplete ontologies) and paraconsistent reasoning (reasoning over inconsistent ontologies), but also improves machine learning in the biomedical domain by incorporating background knowledge from ALC ontologies.

Customized Conversational Recommender Systems

Conversational recommender systems (CRS) aim to capture user's current intentions and provide recommendations through real-time multi-turn conversational interactions. As a human-machine interactive system, it is essential for CRS to improve the user experience. However, most CRS methods neglect the importance of user experience. In this paper, we propose two key points for CRS to improve the user experience: (1) Speaking like a human, human can speak with different styles according to the current dialogue context. (2) Identifying fine-grained intentions, even for the same utterance, different users have diverse finegrained intentions, which are related to users' inherent preference. Based on the observations, we propose a novel CRS model, coined Customized Conversational Recommender System (CCRS), which customizes CRS model for users from three perspectives. For human-like dialogue services, we propose multi-style dialogue response generator which selects context-aware speaking style for utterance generation. To provide personalized recommendations, we extract user's current fine-grained intentions from dialogue context with the guidance of user's inherent preferences. Finally, to customize the model parameters for each user, we train the model from the meta-learning perspective. Extensive experiments and a series of analyses have shown the superiority of our CCRS on both the recommendation and dialogue services.

Joint Abductive and Inductive Neural Logical Reasoning

Neural logical reasoning (NLR) is a fundamental task in knowledge discovery and artificial intelligence. NLR aims at answering multi-hop queries with logical operations on structured knowledge bases based on distributed representations of queries and answers. While previous neural logical reasoners can give specific entity-level answers, i.e., perform inductive reasoning from the perspective of logic theory, they are not able to provide descriptive concept-level answers, i.e., perform abductive reasoning, where each concept is a summary of a set of entities. In particular, the abductive reasoning task attempts to infer the explanations of each query with descriptive concepts, which make answers comprehensible to users and is of great usefulness in the field of applied ontology. In this work, we formulate the problem of the joint abductive and inductive neural logical reasoning (AI-NLR), solving which needs to address challenges in incorporating, representing, and operating on concepts. We propose an original solution named ABIN for AI-NLR. Firstly, we incorporate description logic-based ontological axioms to provide the source of concepts. Then, we represent concepts and queries as fuzzy sets, i.e., sets whose elements have degrees of membership, to bridge concepts and queries with entities. Moreover, we design operators involving concepts on top of the fuzzy set representation of concepts and queries for optimization and inference. Extensive experimental results on two real-world datasets demonstrate the effectiveness of ABIN for AI-NLR.