A hallmark of learning is generalization to novel instances. In speech, exposure to atypical pronunciation drives perceptual adjustment that can generalize to unheard tokens. Prior work has attributed constraints on generalization primarily to acoustic similarity between exposure and test contexts. We propose that generalization can also be understood as an inference problem: listeners must determine whether, and how strongly, a learned phonetic mapping should apply in a new context. We test this proposal using data from a recent experiment in which listeners were exposed to shifted vowel pronunciations and then tested on minimal pairs varying in lexical frequency. Learning effects appeared strongest when the exposure direction aligned with a high-frequency alternative in mixed-frequency pairs, and were absent for low-frequency pairs. The observed pattern could reflect token-level acoustic similarity, reliance on prior expectations, or frequency-dependent constraints in applying the learned mapping. We formalized these alternatives within a Bayesian belief-updating framework: a talker-specific model assuming full transfer, a mixture-of-expectations model that interpolates between the updated representation and the listener’s prior, and a hierarchical Bayesian model that deploys the updated representation with uncertainty. The talker-specific model captured most generalization patterns through its sensitivity to token-level acoustic properties, but overpredicted learning for low-frequency pairs. The hierarchical model best recovered the theoretically central exposure-control contrast pattern, suggesting that lexical frequency may constrain how learned representations are applied. Our results provide a computationally explicit framework for studying how contextual factors shape generalization in speech perception.

Human listeners rapidly adapt to unfamiliar talkers, but the underlying computational mechanisms remain contested. Three candidate hypotheses—pre-linguistic normalization, changes in phonetic category representations, and changing decision biases—have largely been pursued in separation, using subfield-specific paradigms. Researchers working in these paradigms often assume that adaptivity observed in their particular paradigm can only be explained by one of the three mechanisms. We test this assumption for one of the most popular experimental paradigms (lexically-guided perceptual learning or LGPL) using a unified computational framework (ASP). We apply ASP to the largest existing LGPL data: 89,600 categorization responses from over 1000 listeners after lexically-guided exposure to 32 different stimulus sets. Despite the unprecedented scale of these data, we find that behavioral data are equally compatible with all three candidate mechanisms. We discuss how model-guided stimulus selection can increase the diagnosticity of future LGPL experiments. Our simulation code can easily be adapted to other experimental paradigms.

LLM-powered systems require complex multi-step decision-making abilities to solve real-world tasks, yet current planning approaches face a trade-off between the high latency of inference-time search and the limited generalization of supervised fine-tuning. To address this limitation, we introduce SGA-MCTS, a framework that casts LLM planning as non-parametric retrieval. Offline, we leverage Monte Carlo Tree Search (MCTS) to explore the solution space and distill high-fidelity trajectories into State-Goal-Action (SGA) atoms. These atoms are de-lexicalized primitives that abstract concrete entities into symbolic slots, preserving reusable causal logic while discarding domain-specific noise. Online, a retrieval-augmented agent employs a hybrid symbolic-semantic mechanism to fetch relevant SGAs and re-ground them into the current context as soft reasoning hints. Empirical results on complex benchmarks demonstrate that this paradigm enables frozen, open-weights models to match the performance of SOTA systems (e.g., GPT-5) without task-specific fine-tuning. By effectively amortizing the heavy computational cost of search, SGA-MCTS achieves System 2 reasoning depth at System 1 inference speeds, rendering autonomous planning both scalable and real-time feasible.

This paper presents an empirical study to build relation extraction systems in low-resource settings. Based upon recent pre-trained language models, we comprehensively investigate three schemes to evaluate the performance in low-resource settings: (i) different types of prompt-based methods with few-shot labeled data; (ii) diverse balancing methods to address the long-tailed distribution issue; (iii) data augmentation technologies and self-training to generate more labeled in-domain data. We create a benchmark with 8 relation extraction (RE) datasets covering different languages, domains and contexts and perform extensive comparisons over the proposed schemes with combinations. Our experiments illustrate: (i) Though prompt-based tuning is beneficial in low-resource RE, there is still much potential for improvement, especially in extracting relations from cross-sentence contexts with multiple relational triples; (ii) Balancing methods are not always helpful for RE with long-tailed distribution; (iii) Data augmentation complements existing baselines and can bring much performance gain, while self-training may not consistently achieve advancement to low-resource RE. Code and datasets are in https://github.com/zjunlp/LREBench.

We present an open-source and extensible knowledge extraction toolkit DeepKE, supporting complicated low-resource, document-level and multimodal scenarios in the knowledge base population. DeepKE implements various information extraction tasks, including named entity recognition, relation extraction and attribute extraction. With a unified framework, DeepKE allows developers and researchers to customize datasets and models to extract information from unstructured data according to their requirements. Specifically, DeepKE not only provides various functional modules and model implementation for different tasks and scenarios but also organizes all components by consistent frameworks to maintain sufficient modularity and extensibility. We release the source code at GitHub in https://github.com/zjunlp/DeepKE with Google Colab tutorials and comprehensive documents for beginners. Besides, we present an online system in http://deepke.openkg.cn/EN/re_doc_show.html for real-time extraction of various tasks, and a demo video.

This paper presents our systems for the three Subtasks of SemEval Task4: Reading Comprehension of Abstract Meaning (ReCAM). We explain the algorithms used to learn our models and the process of tuning the algorithms and selecting the best model. Inspired by the similarity of the ReCAM task and the language pre-training, we propose a simple yet effective technology, namely, negative augmentation with language model. Evaluation results demonstrate the effectiveness of our proposed approach. Our models achieve the 4th rank on both official test sets of Subtask 1 and Subtask 2 with an accuracy of 87.9% and an accuracy of 92.8%, respectively. We further conduct comprehensive model analysis and observe interesting error cases, which may promote future researches. The code and dataset used in our paper can be found at https://github.com/CheaSim/SemEval2021. The leaderboard can be found at https://competitions.codalab.org/competitions/26153.