Quantization, distillation, pruning and serving work aimed at the same accuracy for less memory, latency and money.
The paper proposes a new algorithm for generating Semantic IDs that are both unique and preserve the structure of the original embedding space. By employing bottom‑up clustering, the method maintains local structure, leading to higher clustering quality. This improved structure enhances the utility of the Semantic IDs for downstream generative retrieval tasks.
ASDA (Automated Skill Distillation and Adaptation) is a framework that improves large language models on financial reasoning tasks without fine‑tuning. It works by having a teacher model analyze a student’s failures, cluster errors, and generate structured skill artifacts—reasoning procedures, code templates, and worked examples—that are injected during inference. On the FAMMA benchmark, ASDA boosts arithmetic reasoning by up to 17.33% and non‑arithmetic reasoning by 5.95%, outperforming existing training‑free methods.
The paper presents a second‑order theory of quantization noise for matrix multiplication, characterizing quantization formats by the variance they assign to each element. It derives a closed‑form signal‑to‑noise‑ratio law for floating‑point rounding, introduces an upper bound κ* that cannot be exceeded by any function‑preserving linear transform, and proposes KBBQ—a method that parameterizes how closely a transform approaches this bound. Experiments on W4A4 across four base models and two FP4 formats show that KBBQ outperforms the previous state of the art without extra deployment‑time computation.
The paper investigates online linear regression with sparse comparators, focusing on feature priming techniques that reweight features using past data. It establishes sparse‑regret lower bounds that invalidate sparse‑logarithmic guarantees for univariate, Pearson, and multivariate priming rules under a past‑only Moore–Penrose protocol, showing ≥Ω(min{T,√d}) clipped regret for unit‑power rules and linear regret for powered rules in high dimensions. The authors also provide tight rank upper bounds for certain priming schemes and present algebraic constructions yielding Ω(min{T,d^{1/4}}) regret for unit‑power multivariate priming, while noting that the exact multivariate frontier remains open.
The paper evaluates entropy-based pruning for compressing Chain-of-Thought (CoT) reasoning in large models. Across multiple models and tasks, low- and high-entropy step selection shows no advantage over random pruning, and low-entropy token retention only helps on mathematical benchmarks due to the low entropy of numeric tokens. Patching a few CoT tokens with their original activations restores near-perfect performance, indicating that task information is distributed throughout the entire reasoning chain rather than concentrated in a small set of tokens.
arXiv:2604.22391v2 Announce Type: replace-cross Abstract: The Super Learner (SL) is a widely used ensemble method that combines point predictions from a library of learners based on their predictive...
arXiv:2609.08638v1 Announce Type: cross Abstract: An action chunk can span several stages of a manipulation task, yet a label for its first step describes only the current stage. We introduce Chunk-A...
The paper introduces TokCode, a token encoding framework that enhances robustness in generative semantic communication by restructuring redundancy in the semantic domain. TokCode leverages a lightweight adapter to transform a large language model into a token encoder, avoiding the need for a dedicated deep model. A channel-quality-aware distillation method (CADET) trains the adapter across diverse erasure rates, producing a reconfigurable low‑rank adapter that enables efficient reinforcement learning and achieves significant improvements in image similarity over existing receiver‑side recovery benchmarks.
arXiv:2605.27156v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, particularly for long-tail do...
arXiv:2605.10889v2 Announce Type: replace-cross Abstract: On-policy distillation offers dense, per-token supervision for training reasoning models; however, it remains unclear under which conditions...
The paper evaluates the scalability and adversarial generalization of Natural Language Inference (NLI) models trained with Group Relative Policy Optimization (GRPO) for Chain-of-Thought learning. By fine‑tuning 7B, 14B, and 32B language models with LoRA and QLoRA, the authors show strong performance on standard and adversarial NLI benchmarks, with the 32B model outperforming supervised baselines on adversarial sets. Using AWQ quantization, the 32B model fits within 22 GB of CUDA memory, demonstrating a scalable, practical framework for robust NLI without sacrificing inference quality.
arXiv:2603.04949v2 Announce Type: replace Abstract: As web agents close the gap with humans on benchmarks, one question arises: Do today's agents perform just as well on tomorrow's web? We introduce...
arXiv:2609.05881v1 Announce Type: cross Abstract: Developers increasingly run large language models locally by pulling quantized GGUF artifacts from public registries, yet nothing in the distribution...
The paper introduces MiLAAP, an attention‑based framework that predicts channel occupancy and node motion in channel‑hopping communication systems without exchanging state information. By leveraging self‑attention and multi‑head attention, each node passively observes local channel activity to forecast interference patterns and mobility, achieving near‑perfect prediction accuracy across varied mobility scenarios and demonstrating zero‑shot generalizability to new channel‑sequence periods.
A*-Thought-V2 is a framework that models Chain-of-Thought reasoning as a geometric trajectory in a 3D PCA space, using explicit-implicit latent tokens to compress steps that deviate from the main question-to-solution direction. The method measures alignment angles to decide which steps remain text and which become latent, and introduces stepwise embedding forcing and label forcing to train the architecture. Experiments on Qwen models show up to 2.6% accuracy gains, halved response length, and significant reductions in computation and training time.
The paper introduces Vision-of-Thought (VoT), a framework that inserts a discrete visual-thinking layer between vision‑language models (VLMs) and diffusion transformers (DiTs). Instead of using VLMs solely as text encoders, they act as multimodal planners that generate VoT tokens—high‑level visual plans such as objects and layouts—before pixel rendering. A specialized VoT tokenizer is trained with a closed‑loop objective combining VLM alignment, feature reconstruction, and vector‑quantization losses, ensuring the tokens are semantically readable by the VLM while preserving necessary visual information. Experimental results show that VoT improves semantic alignment and offers a structured, interpretable interface for controllable generation.
The paper introduces Causal Visual Recurrent Reasoning (CVRR), a method that forces multimodal models to rely on latent visual states by using recurrent computation as the sole image‑conditioned path to prediction. CVRR initializes recurrence from the question hidden state after a pretrained vision‑language model has processed the image, repeatedly updates this state while re‑reading the same visual evidence, and removes all other visual traces before decoding. Experiments on several benchmarks show that CVRR maintains strong performance while other latent reasoners lose visual competence, and causal interventions confirm that predictions depend on the recurrent visual trajectory.
arXiv:2506.12809v2 Announce Type: replace Abstract: The long horizon forecasting (LHF) problem has come up in the time series literature for over the last 35 years or so. This review covers aspects o...
arXiv:2609.07408v1 Announce Type: new Abstract: This study investigates the classification of individuals as healthy or at risk of Parkinson's disease using machine learning (ML) models, focusing on...
arXiv:2609.08123v1 Announce Type: cross Abstract: A robot that can be taught a new task from a handful of demonstrations has to work out for itself what it still cannot do, and then ask for exactly t...