paper-with-me

홈 › Papers

Density Field State Space Models: 1-Bit Distillation, Efficient Inference, and Knowledge Organization in Mamba-2

2026-04-28 · Chirag Shinde arxiv

We present Density Field State Space Models (DF-SSM), a framework for compressing SSMs to a 1-bit scaffold with int8 low-rank correction. Applied to Mamba-2 1.3B, we achieve a 278 MB model (9.7x smaller than the 2.7 GB FP16 teacher) that runs at 21.4x faster inference on GPU (batch=1, relative to the mamba-ssm reference implementation) while maintaining downstream task performance within 2-4 percentage points of BitMamba-2, a 1.58-bit model trained from scratch on 150B tokens. The distillation itself requires only 32M tokens and 6 hours on a single A100 GPU, though it presupposes a pretrained FP16 teacher. We develop an optimized inference pipeline combining cuBLAS INT8 tensor cores for the scaffold matmul, custom CUDA kernels for stateful SSM and convolution operations, and an AVX-512 CPU backend for efficient deployment on both GPU and CPU. Beyond compression, we investigate the internal knowledge organization of the resulting model, discovering three distinct processing phases: intent classification (layers 0-3, operating in an abstract space with no vocabulary alignment), knowledge retrieval (layers 25-35, where factual associations localize to a 5-layer window), and output formatting (layers 36-47, where category structure dissolves). Through systematic analysis of 445 factual prompts across 19 categories, we find that early-layer classification is syntactic (driven by template structure) rather than semantic, and that the model exhibits well-organized knowledge representations despite weak factual recall--suggesting that representational structure may precede factual strength.

📄 PDF Abstract BibTeX arXiv:2606.10932

Code (0)

등록된 구현이 없습니다.

Tasks

Intent Classification

Similar Papers 제목 키워드 기반

Bridging the Gap Between Variational Inference and Wasserstein Gradient Flows

2023-10-31 · Mingxuan Yi, Song Liu

Variational inference is a technique that approximates a target distribution by optimizing within the parameter space of variational families. On the other hand, Wasserstein gradient flows describe optimization within th…

Variational Inference

GAGS: Granularity-Aware Feature Distillation for Language Gaussian Splatting

2024-12-18 · Yuning Peng, Haiping Wang, YuAn Liu, Chenglu Wen 외

3D open-vocabulary scene understanding, which accurately perceives complex semantic properties of objects in space, has gained significant attention in recent years. In this paper, we propose GAGS, a framework that disti…

Scene UnderstandingSemantic SegmentationVisual Grounding

One-Step Diffusion with Inverse Residual Fields for Unsupervised Industrial Anomaly Detection

2026-04-20 · Boan Zhang, Wen Li, Guanhua Yu, Xiyang Liu 외 arxiv

Diffusion models have achieved outstanding performance in unsupervised industrial anomaly detection (uIAD) by learning a manifold of normal data under the common assumption that off-manifold anomalies are harder to gener…

Anomaly Detection

Boosting Self-Supervision for Single-View Scene Completion via Knowledge Distillation

2024-04-11 · CVPR 2024 1 · Keonhee Han, Dominik Muhle, Felix Wimbauer, Daniel Cremers

Inferring scene geometry from images via Structure from Motion is a long-standing and fundamental problem in computer vision. While classical approaches and, more recently, depth map predictions only focus on the visible…

Depth EstimationDepth PredictionKnowledge DistillationNovel View Synthesis

Few-step Flow for 3D Generation via Marginal-Data Transport Distillation

2025-09-04 · Zanwei Zhou, Taoran Yi, Jiemin Fang, Chen Yang 외 arxiv

Flow-based 3D generation models typically require dozens of sampling steps during inference. Though few-step distillation methods, particularly Consistency Models (CMs), have achieved substantial advancements in accelera…

3D Generation