arXiv:2605.13375v1 Announce Type: cross Abstract: In Vision-Language Models (VLMs), processing a massive number of visual tokens incurs prohibitive computational overhead. While recent training-aware pruning methods attempt to selectively discard redundant tokens, they largely rely on continuous-gradient relaxations. However, visual token pruning is inherently a discrete, non-convex combinatorial problem; consequently, these continuous approximations frequently trap the […]
GAAMA: Graph Augmented Associative Memory for Agents
arXiv:2603.27910v2 Announce Type: replace Abstract: AI agents that interact with users across multiple sessions require persistent long-term memory to maintain coherent, personalized behavior. Current approaches either rely on flat retrieval-augmented generation (RAG), which loses structural relationships among memories, or use entity-centric knowledge graphs that suffer from mega-hub effects in conversational data, diluting graph-based relevance propagation. […]
Combining Mechanical and Agentic Specification Inference for Move
arXiv:2605.10005v2 Announce Type: replace-cross Abstract: In this paper, we describe early work on a specification inference tool for the Move Prover that combines a weakest-precondition (WP) analysis over Move bytecode with an agentic coding CLI such as Claude Code. Specification inference reduces the boilerplate of writing specifications in Move: in order to verify a high-level […]
BEAVER: An Enterprise Benchmark for Text-to-SQL
arXiv:2409.02038v3 Announce Type: replace-cross Abstract: Existing text-to-SQL benchmarks have largely been constructed from public databases with well-structured schemas and simplistic question-SQL pairs. While large language models (LLMs) excel on these settings, their efficacy in complex private enterprise environments, characterized by intricate schemas, domain knowledge, and analytical user queries involving sophisticated structures and functions, remains unproven. […]
Latent-Augmented Discrete Diffusion Models
arXiv:2510.18114v3 Announce Type: replace-cross Abstract: Discrete diffusion models have emerged as a powerful class of models and a promising route to fast language generation, but practical implementations typically rely on factored reverse transitions ignoring cross-token dependencies and degrading few-step performance. We propose Latent-Augmented Discrete Diffusion (LADD), which introduces a learnable auxiliary latent channel and performs […]
Efficient compression of neural networks and datasets
arXiv:2505.17469v2 Announce Type: replace-cross Abstract: Compression and generalization are fundamentally related through Solomonoff induction and the minimum description length principle (MDL), which predict that simpler models generalize better when data arises from low-complexity distributions. In this article, we combine insights from algorithmic information theory and techniques from neural network pruning to improve model generalization by […]
A Horn extension of DL-Lite with NL data complexity
arXiv:2605.13367v1 Announce Type: cross Abstract: The literature on ontology-mediated query answering (OMQA) has been shaped by two key results: first-order rewritability for DL-Lite, and PTime-hardness of data complexity for essentially every description logic beyond it. This has effectively positioned DL-Lite as the only practical choice for query rewriting, restricting OMQA solutions to first-order queries and […]
FOAM: Blocked State Folding for Memory-Efficient LLM Training
arXiv:2512.07112v2 Announce Type: replace-cross Abstract: Large language models (LLMs) have demonstrated remarkable performance due to their large parameter counts and extensive training data. However, their scale leads to significant memory bottlenecks during training, especially when using memory-intensive optimizers like Adam. Existing memory-efficient approaches often rely on techniques such as singular value decomposition (SVD), projections, or […]
MCLR: Improving Conditional Modeling via Inter-Class Likelihood-Ratio Maximization and Unifying Classifier-Free Guidance with Alignment Objectives
arXiv:2603.22364v3 Announce Type: replace-cross Abstract: Diffusion models achieve strong performance in generative modeling, but their success often relies heavily on classifier-free guidance (CFG), an inference-time heuristic that modifies the sampling trajectory. In theory, diffusion models trained with standard denoising score matching (DSM) should recover the target data distribution, raising two fundamental questions: (i) why is […]
Does Your Neural Network Extrapolate? Feature Engineering as Identifiability Bias for OOD Generalization
arXiv:2605.07483v2 Announce Type: replace-cross Abstract: Successful deep neural networks discover salient features of data. We show when and why they fail to learn out-of-distribution (OOD)-relevant representations from an in-distribution (ID) training window. This requires decoupling feature learning from data-generating-process (DGP) identifiability. From a single training window, OOD extrapolation is non-identifiable: infinitely many DGPs are $varepsilon$-observationally […]
AI Harness Engineering: A Runtime Substrate for Foundation-Model Software Agents
arXiv:2605.13357v1 Announce Type: cross Abstract: Foundation models have transformed automated code generation, yet autonomous software-engineering agents remain unreliable in realistic development settings. The dominant explanation locates this gap in model capability. We propose a different locus: software-engineering capability emerges from a model-harness-environment system, in which a runtime substrate — the harness — mediates how a […]
Identifying AI Web Scrapers Using Canary Tokens
arXiv:2605.13706v1 Announce Type: cross Abstract: From pre-training to query-time augmentation, web-scraped data helps to improve the quality and contextual relevancy of content generated by large language models (LLMs). However, large-scale web scraping to feed LLMs can affect site stability and raise legal, privacy, or ethics concerns. If website owners wish to limit LLM-related web scraping […]