State-of-the-art symbolic music generation models have recently achieved remarkable output quality, yet explicit control over compositional features, such as tonal tension, remains challenging. We propose a novel approach that integrates a computational tonal tension model, based on tonal interval vector analysis, into a Transformer framework. Our method employs a two-level beam search strategy during inference. At the token level, generated candidates are re-ranked using model probability and diversity metrics to maintain overall quality. At the bar level, a tension-based re-ranking is applied to ensure that the generated music aligns with a desired tension curve. Objective evaluations indicate that our approach effectively modulates tonal tension, and subjective listening tests confirm that the system produces outputs that align with the target tension. These results demonstrate that explicit tension conditioning through a dual-level beam search provides a powerful and intuitive tool to guide AI-generated music. Furthermore, our experiments demonstrate that our method can generate multiple distinct musical interpretations under the same tension condition.

The widespread adoption of Large Language Models (LLMs) as re-rankers is shifting recommender systems towards a user-centric paradigm. However, a significant gap remains: current re-rankers often lack mechanisms for fine-grained user control. They struggle to balance inherent user preferences with multiple attribute-based constraints, often resorting to simplistic hard filtering that can excessively narrow the recommendation pool and yield suboptimal results. This limitation leaves users as passive recipients rather than active collaborators in the recommendation process. To bridge this gap, we propose COREC, a novel token-augmented re-ranking framework that incorporates specific user requirements in co-creating the recommendation outcome. COREC empowers users to steer re-ranking results with precise and flexible control via explicit, attribute-based signals. The framework learns to balance these commands against latent preferences, yielding rankings that adhere to user instructions without sacrificing personalization. Experiments show that COREC: (1) exceeds state-of-the-art baselines on standard recommendation effectiveness and (2) demonstrates superior adherence to specific attribute requirements, proving that COREC enables fine-grained and predictable manipulation of the rankings.

We present UTF8Tokenizer, a minimalist byte-level tokenizer that maps text exactly to IDs corresponding to the bytes underlying the text's UTF-8 encoding (e.g., byte x09 is token ID 9). Unlike prior byte-level approaches (Xue et al., 2021; Pagnoni et al., 2025), our implementation never introduces out-of-range IDs (i.e. there is no token ID 256) or auxiliary tokens: all special behavior (e.g., padding, boundaries, conversation structure, attention segments, tool calling, "thinking" spans, etc.) is encoded using C0 control bytes - just as ASCII was originally designed to embed control information alongside printable text. These design principles yield practical benefits: (1) faster tokenization (14x) and significantly lower host-device transfer (8x less than int64); (2) simple, shareable 256*d embedding tables that can be aligned across models; and (3) a training-time enhancement via bit-biased embeddings, which exposes per-byte bit structure and can be added to the embedding table post-training, removing inference costs. Our HuggingFace-compatible implementation improves language modeling convergence.

Recent advancements in Large Language Models (LLMs) have leveraged increased test-time computation to enhance reasoning capabilities, a strategy that, while effective, incurs significant latency and resource costs, limiting their applicability in real-world time-constrained or cost-sensitive scenarios. This paper introduces BudgetThinker, a novel framework designed to empower LLMs with budget-aware reasoning, enabling precise control over the length of their thought processes. We propose a methodology that periodically inserts special control tokens during inference to continuously inform the model of its remaining token budget. This approach is coupled with a comprehensive two-stage training pipeline, beginning with Supervised Fine-Tuning (SFT) to familiarize the model with budget constraints, followed by a curriculum-based Reinforcement Learning (RL) phase that utilizes a length-aware reward function to optimize for both accuracy and budget adherence. We demonstrate that BudgetThinker significantly surpasses strong baselines in maintaining performance across a variety of reasoning budgets on challenging mathematical benchmarks. Our method provides a scalable and effective solution for developing efficient and controllable LLM reasoning, making advanced models more practical for deployment in resource-constrained and real-time environments.

The rapid progress seen in terms of large-scale generative AI is largely based on the attention mechanism. It is conversely non-trivial to conceive small-scale applications for which attention-based architectures outperform traditional approaches, such as multi-layer perceptrons or recurrent networks. We examine this problem in the context of 'task switching'. In this framework models work on ongoing token sequences with the current task being determined by stochastically interspersed control tokens. We show that standard transformers cannot solve a basic task switching reference model based on finite domain arithmetics which contains subtasks dedicated to increment / addition / reverse copy / context (IARC). We show that transformers, long short-term memory recurrent networks (LSTM), and plain multi-layer perceptrons (MLPs) achieve similar, but only modest prediction accuracies. We enlarge our comparative study by including an extension of the standard transformer architecture to its non-translational invariant counterpart, the cisformer, and an alternative attention mechanism, extensive attention. A combination of the latter is found to be the only model able to achieve considerable performance levels, of around 95%. Our results indicate that the workings of attention can be understood better, and even improved, when comparing qualitatively different formulations in task-switching settings.

Prior models and benchmarks focus on closed-loop motion simulation for initial agents in a scene. This is problematic for long-term simulation. Agents enter and exit the scene as the ego vehicle enters new regions. W e propose InfGen, a unified next-token prediction model that performs interleaved closed-loop motion simulation and scene generation. InfGen automatically switches between closed-loop motion simulation and scene generation mode. It enables stable long-term rollout simulation. InfGen performs at the state-of-the-art in short-term (9s) traffic simulation, and significantly outperforms all other methods in long-term (30s) simulation.

Reasoning Language Models, capable of extended chain-of-thought reasoning, have demonstrated remarkable performance on tasks requiring complex logical inference. However, applying elaborate reasoning for all queries often results in substantial computational inefficiencies, particularly when many problems admit straightforward solutions. This motivates an open question: Can LLMs learn when to think? To answer this, we propose Thinkless, a learnable framework that empowers an LLM to adaptively select between short-form and long-form reasoning, based on both task complexity and the model's ability. Thinkless is trained under a reinforcement learning paradigm and employs two control tokens, for concise responses and for detailed reasoning. At the core of our method is a Decoupled Group Relative Policy Optimization (DeGRPO) algorithm, which decomposes the learning objective of hybrid reasoning into two components: (1) a control token loss that governs the selection of the reasoning mode, and (2) a response loss that improves the accuracy of the generated answers. This decoupled formulation enables fine-grained control over the contributions of each objective, stabilizing training and effectively preventing collapse observed in vanilla GRPO. Empirically, on several benchmarks such as Minerva Algebra, MATH-500, and GSM8K, Thinkless is able to reduce the usage of long-chain thinking by 50% - 90%, significantly improving the efficiency of Reasoning Language Models. The code is available at https://github.com/VainF/Thinkless

Achieving full-duplex communication in spoken dialogue systems (SDS) requires real-time coordination between listening, speaking, and thinking. This paper proposes a semantic voice activity detection (VAD) module as a dialogue manager (DM) to efficiently manage turn-taking in full-duplex SDS. Implemented as a lightweight (0.5B) LLM fine-tuned on full-duplex conversation data, the semantic VAD predicts four control tokens to regulate turn-switching and turn-keeping, distinguishing between intentional and unintentional barge-ins while detecting query completion for handling user pauses and hesitations. By processing input speech in short intervals, the semantic VAD enables real-time decision-making, while the core dialogue engine (CDE) is only activated for response generation, reducing computational overhead. This design allows independent DM optimization without retraining the CDE, balancing interaction accuracy and inference efficiency for scalable, next-generation full-duplex SDS.

We introduce Composer's Assistant 2, a system for interactive human-computer composition in the REAPER digital audio workstation. Our work upgrades the Composer's Assistant system (which performs multi-track infilling of symbolic music at the track-measure level) with a wide range of new controls to give users fine-grained control over the system's outputs. Controls introduced in this work include two types of rhythmic conditioning controls, horizontal and vertical note onset density controls, several types of pitch controls, and a rhythmic interest control. We train a T5-like transformer model to implement these controls and to serve as the backbone of our system. With these controls, we achieve a dramatic improvement in objective metrics over the original system. We also study how well our model understands the meaning of our controls, and we conduct a listening study that does not find a significant difference between real music and music composed in a co-creative fashion with our system. We release our complete system, consisting of source code, pretrained models, and REAPER scripts.

This study addresses the hallucination problem in large language models (LLMs). We adopted Retrieval-Augmented Generation(RAG) (Lewis et al., 2020), a technique that involves embedding relevant information in the prompt to obtain accurate answers. However, RAG also faced inherent issues in retrieving correct information. To address this, we employed the Dense Passage Retrieval(DPR) (Karpukhin et al., 2020) model for fetching domain-specific documents related to user queries. Despite this, the DPR model still lacked accuracy in document retrieval. We enhanced the DPR model by incorporating control tokens, achieving significantly superior performance over the standard DPR model, with a 13% improvement in Top-1 accuracy and a 4% improvement in Top-20 accuracy.