Quantization, distillation, pruning and serving work aimed at the same accuracy for less memory, latency and money.
Diffusion Transformers (DiTs) incur quadratic self-attention cost over spatiotemporal tokens. Existing training-free sparse attention methods often construct sparse masks from block-level or cluster-level proxy scores, which can obscure fine-grained differences among keys and miss high contribution keys under aggressive sparsity.
Fashion retrieval often requires satisfying multiple attributes at once, such as category, color, pattern, and demographic. Monolithic embeddings mix these signals into a single vector, making attribute-specific control difficult at retrieval time.
Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods. In this paper, we ask whether such transfer can instead occur at test time.
Use this plain-text version for the arXiv abstract field: Learned image compression (LIC) models achieve strong rate-distortion performance but are hindered by high computational complexity and encoding-decoding mismatches across heterogeneous hardware platforms. Uniform fixed-precision quantization alleviates these issues but suffers severe quality degradation at low bit widths because it ignores differences in the quantization sensitivities of individual layers.
Boosted decision trees (BDTs) are widely used in latency-critical applications, but efficient hardware deployment remains challenging. Existing designs often rely on uniform or manually tuned fixed-point formats, which can introduce unnecessary hardware cost or accuracy loss.
Small Language Models (SLMs) have emerged as a more efficient alternative to traditional Large Language Models (LLMs), offering promising potential in resource-constrained scenarios. Existing approaches to building SLMs typically follow two paths: training compact models from scratch, or compressing larger pre-trained models using methods such as pruning, quantization, or distillation.
Aggressive quantization disproportionately harms multilingual capability: in the sub-4B INT3 GPTQ regime, we measure 2-4x larger perplexity degradation on non-English languages than on English. We propose Language-Conditional Dequantization (LCD), a post-hoc method that attaches per-language rank-2 LoRA corrections to the linear layers of an already-quantized model, adding 0.
Diffusion Large Language Models (dLLMs) have emerged as a competitive alternative to autoregressive language models, offering the potential for substantially faster inference through parallel decoding. Existing parallel decoding schedulers typically commit positions only after they meet a per-position criterion, overlooking how early commitments may benefit subsequent decoding.
arXiv:2605. 28803v3 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models unify perception, reasoning, and control in a single policy, but their multi-billion-parameter backbones and diffusion-based action heads make on-device deployment prohibitively expensive.
arXiv:2608. 10484v1 Announce Type: cross Abstract: Action verbs describe not only the physical outcomes of actions, but also how those actions are performed.
arXiv:2608. 10392v1 Announce Type: new Abstract: Mixture-of-experts (MoE) models have recently moved beyond routing a fixed number of complete experts.
arXiv:2608. 10523v1 Announce Type: cross Abstract: \texttt{TensorSketch} by~\cite{pham2013fast,kar2012random} provides efficient sketching algorithms for high-dimensional polynomial kernels $\vec{x}^{\otimes p} \in \R^{d^p}$.
arXiv:2608. 11045v1 Announce Type: new Abstract: ReRound (Reconstructive Rounding) is a post-training quantization method that addresses the midpoint ambiguity inherent in standard round-to-nearest (RTN) schemes when quantizing weights near the centers of quantization intervals.
arXiv:2608. 10526v1 Announce Type: cross Abstract: Motivated by decentralized applications, we study cooperative multi-agent bandits in continuous (Lipschitz) action spaces when the Lipschitz constant is unknown.
arXiv:2608. 10812v1 Announce Type: cross Abstract: We study reference-free post-training for multilingual machine translation with open large language models.
arXiv:2605. 08731v3 Announce Type: replace-cross Abstract: A JPEG decoder benchmark can combine worker counts, CPUs, and datasets in one large result matrix.
arXiv:2608. 10430v1 Announce Type: cross Abstract: Large Language Models (LLMs) deployed as AI agents frequently exhibit user specification-grounding failures, executing hallucinated, undesired actions to force a resolution rather than expressing uncertainty.
arXiv:2608. 10805v1 Announce Type: cross Abstract: Wavelet convolution (WTConv) has emerged as an increasingly popular drop-in replacement for standard convolutions, expanding a network's receptive field exponentially with the number of decomposition levels while keeping the parameter count linear.
arXiv:2608. 10524v1 Announce Type: cross Abstract: Driven by the rapid advancement of vision-language representation learning, Text-based Image Retrieval (TBIR) has made notable progress.
arXiv:2608. 10402v1 Announce Type: new Abstract: Reinforcement learning (RL) for large language models is moving toward multi-turn agentic workloads, where rollout tasks repeatedly pause for external environments, resume with growing contexts, and finish at highly variable times.