Modern language models are powerful, but typically static after deployment. A major obstacle to building models that continually learn over time is catastrophic forgetting, where updating on new data erases previously acquired capabilities. Motivated by the intuition that mitigating forgetting is challenging because trainable parameters are shared across all tasks, we investigate whether sparse parameter updates can enable learning without catastrophic forgetting. We introduce sparse memory finetuning, leveraging memory layer models (Berges et al., 2024), which are sparsely updated by design. By updating only the memory slots that are highly activated by a new piece of knowledge relative to usage on pretraining data, we reduce interference between new knowledge and the model's existing capabilities. We evaluate learning and forgetting compared to full finetuning and parameter-efficient finetuning with LoRA on two question answering tasks. We find that sparse memory finetuning learns new knowledge while exhibiting substantially less forgetting: while NaturalQuestions F1 drops by 89% after full finetuning on new facts and 71% with LoRA, sparse memory finetuning yields only an 11% drop with the same level of new knowledge acquisition. Our results suggest sparsity in memory layers offers a promising path toward continual learning in large language models.

The developments in Generative AI technologies have paved the way for numerous innovations in different fields. Recently, Generative AI has been proposed as a competitor to AES systems in evaluating student essays automatically. Considering the potential limitations of AI in processing idioms, this study assessed the scoring performances of Generative AI models for essays with and without idioms by incorporating insights from Corpus Linguistics and Computational Linguistics. Two equal essay lists were created from 348 student essays taken from a corpus: one with multiple idioms present in each essay and another with no idioms in essays. Three Generative AI models (ChatGPT, Gemini, and Deepseek) were asked to score all essays in both lists three times, using the same rubric used by human raters in assigning essay scores. The results revealed excellent consistency for all models, but Gemini outperformed its competitors in interrater reliability with human raters. There was also no detectable bias for any demographic group in AI assessment. For essays with multiple idioms, Gemini followed a the most similar pattern to human raters. While the models in the study demonstrated potential for a hybrid approach, Gemini was the best candidate for the task due to its ability to handle figurative language and showed promise for handling essay-scoring tasks alone in the future.

Reward machines are an established tool for dealing with reinforcement learning problems in which rewards are sparse and depend on complex sequences of actions. However, existing algorithms for learning reward machines assume an overly idealized setting where rewards have to be free of noise. To overcome this practical limitation, we introduce a novel type of reward machines, called stochastic reward machines, and an algorithm for learning them. Our algorithm, based on constraint solving, learns minimal stochastic reward machines from the explorations of a reinforcement learning agent. This algorithm can easily be paired with existing reinforcement learning algorithms for reward machines and guarantees to converge to an optimal policy in the limit. We demonstrate the effectiveness of our algorithm in two case studies and show that it outperforms both existing methods and a naive approach for handling noisy reward functions.

When working on digital devices, people often face distractions that can lead to a decline in productivity and efficiency, as well as negative psychological and emotional impacts. To address this challenge, we introduce a novel Artificial Intelligence (AI) assistant that elicits a user's intention, assesses whether ongoing activities are in line with that intention, and provides gentle nudges when deviations occur. The system leverages a large language model to analyze screenshots, application titles, and URLs, issuing notifications when behavior diverges from the stated goal. Its detection accuracy is refined through initial clarification dialogues and continuous user feedback. In a three-week, within-subjects field deployment with 22 participants, we compared our assistant to both a rule-based intent reminder system and a passive baseline that only logged activity. Results indicate that our AI assistant effectively supports users in maintaining focus and aligning their digital behavior with their intentions. Our source code is publicly available at https://intentassistant.github.io

Federated learning promises to revolutionize machine learning by enabling collaborative model training without compromising data privacy. However, practical adaptability can be limited by critical factors, such as the participation dilemma. Participating entities are often unwilling to contribute to a learning system unless they receive some benefits, or they may pretend to participate and free-ride on others. This chapter identifies the fundamental challenges in designing incentive mechanisms for federated learning systems. It examines how foundational concepts from economics and game theory can be applied to federated learning, alongside technology-driven solutions such as blockchain and deep reinforcement learning. This work presents a comprehensive taxonomy that thoroughly covers both centralized and decentralized architectures based on the aforementioned theoretical concepts. Furthermore, the concepts described are presented from an application perspective, covering emerging industrial applications, including healthcare, smart infrastructure, vehicular networks, and blockchain-based decentralized systems. Through this exploration, this chapter demonstrates that well-designed incentive mechanisms are not merely optional features but essential components for the practical success of federated learning. This analysis reveals both the promising solutions that have emerged and the significant challenges that remain in building truly sustainable, fair, and robust federated learning ecosystems.

Monte Carlo (TSMC) to provide scalable stepwise supervision using small LMs. This yields more efficient training, improved stability, and higher accuracy. OpenAI's o1 series (OpenAI, 2024) introduce inference-time scaling by increasing the length of the Chain-of-Thought (CoT) (Wei et al., 2022) reasoning process. Despite their empirical success, RL approaches that generate the entire reasoning chain in a single forward pass face notable limitations, including CoT derailment, where the reasoning trajectory drifts off course due to accumulated errors, and the inherent challenges of long-horizon RL with sparse outcome rewards. This sequential scaling strategy, i.e., simply extending the CoT length, can therefore be insufficient (Y ang et al., 2025). To improve planning quality, we introduce Multi-Path Plan Aggregation (MPP A). For each planning step, the model generates multiple alternative plans and aggregates them into an improved plan before proceeding to the subsequent execution steps. Beyond enhancing planning, we identify a fundamental challenge in credit assignment for long-horizon policy learning (Kaelbling et al., 1996). Existing RL fine-tuning frameworks struggle to provide effective process-level supervision (Guo et al., 2025). First, evaluating the correctness of intermediate steps is inherently difficult. Automated annotation using LLM judges (Gu et al., 2024) often yield unreliable or noisy signals Second, introducing a separate process reward model (PRM) adds complexity. We then define the process preference between two candidate continuations at the same step by comparing their incremental log-weights. We repurpose Twisted Sequential Monte Carlo (TSMC) to provide process-level preferences for online Step-DPO training. Results show that our approach consistently outperforms both distillation-based long-CoT methods and RL methods that rely solely on outcome rewards. The Chain-of-Thought trajectories can be lengthy and the positions of the first error vary considerably, making outcome-based RL fine-tuning inefficient. Training long trajectories with outcome rewards is highly inefficient.

Large Language Models (LLMs) are powerful reasoners in natural language, but their actions are typically confined to outputting vocabulary tokens. As a result, interactions with external environments -- such as symbolic operators or simulators -- must be expressed through text in predefined formats, parsed, and routed to external interfaces. This overloads the model's language with both reasoning and control duties, and requires a hand-crafted parser, external to the LLM. To address this, we decouple environment interactions from language by internalizing them in an Expanded Action space (ExpA), beyond the vocabulary. The model starts reasoning in the default language environment, but may trigger routing actions and switch to an external environment at any time. From there, the model can only invoke environment-specific actions, receive feedback from the environment, and potentially route back to language as a result. To promote effective exploration of the expanded action space and new environments, we introduce ExpA Reinforcement Learning (EARL) with counterfactual policy optimization. On tasks requiring multi-turn interactions and contingent planning, EARL outperforms strong baselines with vocabulary-constrained actions. It performs robustly across calculator-based multi-task learning and, in the partially observed sorting problem, achieves perfect Sort-4 accuracy while self-discovering an efficient algorithm competitive with classical designs.

The immense computational cost of training Large Language Models (LLMs) presents a major barrier to innovation. While FP8 training offers a promising solution with significant theoretical efficiency gains, its widespread adoption has been hindered by the lack of a comprehensive, open-source training recipe. To bridge this gap, we introduce an end-to-end FP8 training recipe that seamlessly integrates continual pre-training and supervised fine-tuning. Our methodology employs a fine-grained, hybrid-granularity quantization strategy to maintain numerical fidelity while maximizing computational efficiency. Through extensive experiments, including the continue pre-training of models on a 160B-token corpus, we demonstrate that our recipe is not only remarkably stable but also essentially lossless, achieving performance on par with the BF16 baseline across a suite of reasoning benchmarks. Crucially, this is achieved with substantial efficiency improvements, including up to a 22% reduction in training time, a 14% decrease in peak memory usage, and a 19% increase in throughput. Our results establish FP8 as a practical and robust alternative to BF16, and we will release the accompanying code to further democratize large-scale model training.

This paper introduces a simple and scalable approach to improve the data efficiency of large language model (LLM) training by augmenting existing text data with thinking trajectories. The compute for pre-training LLMs has been growing at an unprecedented rate, while the availability of high-quality data remains limited. Consequently, maximizing the utility of available data constitutes a significant research challenge. A primary impediment is that certain high-quality tokens are difficult to learn given a fixed model capacity, as the underlying rationale for a single token can be exceptionally complex and deep. To address this issue, we propose Thinking augmented Pre-Training (TPT), a universal methodology that augments text with automatically generated thinking trajectories. Such augmentation effectively increases the volume of the training data and makes high-quality tokens more learnable through step-by-step reasoning and decomposition. We apply TPT across diverse training configurations up to $100$B tokens, encompassing pre-training with both constrained and abundant data, as well as mid-training from strong open-source checkpoints. Experimental results indicate that our method substantially improves the performance of LLMs across various model sizes and families. Notably, TPT enhances the data efficiency of LLM pre-training by a factor of $3$. For a $3$B parameter model, it improves the post-training performance by over $10\%$ on several challenging reasoning benchmarks.

Spiking Neural Networks (SNNs) provide an efficient framework for processing dynamic spatio-temporal signals and for investigating the learning principles underlying biological neural systems. A key challenge in training SNNs is to solve both spatial and temporal credit assignment. The dominant approach for training SNNs is Backpropagation Through Time (BPTT) with surrogate gradients. However, BPTT is in stark contrast with the spatial and temporal locality observed in biological neural systems and leads to high computational and memory demands, limiting efficient training strategies and on-device learning. Although existing local learning rules achieve local temporal credit assignment by leveraging eligibility traces, they fail to address the spatial credit assignment without resorting to auxiliary layer-wise matrices, which increase memory overhead and hinder scalability, especially on embedded devices. In this work, we propose Traces Propagation (TP), a forward-only, memory-efficient, scalable, and fully local learning rule that combines eligibility traces with a layer-wise contrastive loss without requiring auxiliary layer-wise matrices. TP outperforms other fully local learning rules on NMNIST and SHD datasets. On more complex datasets such as DVS-GESTURE and DVS-CIFAR10, TP showcases competitive performance and scales effectively to deeper SNN architectures such as VGG-9, while providing favorable memory scaling compared to prior fully local scalable rules, for datasets with a significant number of classes. Finally, we show that TP is well suited for practical fine-tuning tasks, such as keyword spotting on the Google Speech Commands dataset, thus paving the way for efficient learning at the edge.