AIBullisharXiv – CS AI · Mar 167/10
🧠Researchers introduce a novel optimization framework that integrates the Minimum Description Length (MDL) principle directly into deep neural network training dynamics. The method uses geometrically-grounded cognitive manifolds with coupled Ricci flow to create autonomous model simplification while maintaining data fidelity, with theoretical guarantees for convergence and practical O(N log N) complexity.
AIBullisharXiv – CS AI · Mar 67/10
🧠Researchers propose asymmetric transformer attention where keys use fewer dimensions than queries and values, achieving 75% key cache reduction with minimal quality loss. The technique enables 60% more concurrent users for large language models by saving 25GB of KV cache per user for 7B parameter models.
🏢 Perplexity
AIBullisharXiv – CS AI · Mar 56/10
🧠Researchers successfully developed Bielik-Q2-Sharp, the first systematic evaluation of extreme 2-bit quantization for Polish language models, achieving near-baseline performance while significantly reducing model size. The study compared six quantization methods on an 11B parameter model, with the best variant maintaining 71.92% benchmark performance versus 72.07% baseline at just 3.26 GB.
AIBullisharXiv – CS AI · Mar 57/10
🧠Researchers from KAIST propose AMiD, a new knowledge distillation framework that improves the efficiency of training smaller language models by transferring knowledge from larger models. The technique introduces α-mixture assistant distribution to address training instability and capacity gaps in existing approaches.
AIBullisharXiv – CS AI · Mar 46/102
🧠Researchers propose Router Knowledge Distillation (Router KD) to improve retraining-free compression of Mixture-of-Experts (MoE) models by calibrating routers while keeping expert parameters unchanged. The method addresses router-expert mismatch issues that cause performance degradation in compressed MoE models, showing particularly strong results in fine-grained MoE architectures.
AIBullisharXiv – CS AI · Mar 37/102
🧠ButterflyMoE introduces a breakthrough approach to reduce memory requirements for AI expert models by 150× through geometric parameterization instead of storing independent weight matrices. The method uses shared ternary prototypes with learned rotations to achieve sub-linear memory scaling, enabling deployment of multiple experts on edge devices.
AIBullisharXiv – CS AI · Feb 277/106
🧠Researchers developed TT-SEAL, a selective encryption framework for compressed AI models using Tensor-Train Decomposition that maintains security while encrypting only 4.89-15.92% of parameters. The system achieves the same robustness as full encryption while reducing AES decryption overhead in end-to-end latency from 58% to as low as 2.76%.
AIBullisharXiv – CS AI · Feb 277/105
🧠Tencent Hunyuan team introduces AngelSlim, a comprehensive toolkit for large model compression featuring quantization, speculative decoding, and pruning techniques. The toolkit includes the first industrially viable 2-bit large model (HY-1.8B-int2) and achieves 1.8x to 2.0x throughput gains while maintaining output quality.
AIBullisharXiv – CS AI · Feb 277/108
🧠Researchers introduce UniQL, a unified framework for quantizing and compressing large language models to run efficiently on mobile devices. The system achieves 4x-5.7x memory reduction and 2.7x-3.4x speed improvements while maintaining accuracy within 5% of original models.
AIBullisharXiv – CS AI · Feb 277/107
🧠Researchers have developed a unified framework using Spectral Geometry and Random Matrix Theory to address reliability and efficiency challenges in large language models. The study introduces EigenTrack for real-time hallucination detection and RMT-KD for model compression while maintaining accuracy.
AINeutralarXiv – CS AI · Feb 277/106
🧠Researchers establish theoretical foundations for neural network superposition, proving lower bounds that require at least Ω(√m' log m') neurons and Ω(m' log m') parameters to compute m' features. The work demonstrates exponential complexity gaps between computing versus merely representing features and provides first subexponential bounds on network capacity.
AIBullishHugging Face Blog · Sep 187/105
🧠The article discusses techniques for fine-tuning large language models (LLMs) to achieve extreme quantization down to 1.58 bits, making the process more accessible and efficient. This represents a significant advancement in model compression technology that could reduce computational requirements and costs for AI deployment.
AIBullishHugging Face Blog · May 247/108
🧠The article discusses advances in making Large Language Models (LLMs) more accessible through bitsandbytes library, 4-bit quantization techniques, and QLoRA (Quantized Low-Rank Adaptation). These technologies enable running and fine-tuning large AI models on consumer hardware with significantly reduced memory requirements.
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers introduce SCENIC, a neural framework designed to optimize language models for edge IoT devices by enabling them to convert natural language commands into structured smart-home instructions. The system achieves 99% accuracy on benchmarks while reducing model size by 25% through pruning and quantization, addressing the practical challenge of deploying AI on memory-constrained devices.
🏢 Nvidia
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers introduce HyperAdapter, a parameter-efficient fine-tuning method for vision transformers that adapts model weights through hypergraph-structured token groupings rather than individual tokens. The approach demonstrates consistent performance improvements over existing adapter methods while maintaining computational efficiency, suggesting that adaptation space design is critical for vision transformer transfer learning.
AIBullisharXiv – CS AI · Jun 236/10
🧠Researchers introduce PRIDE, a knowledge distillation method that compresses large language models for empathetic dialogue while maintaining quality through privileged information available only during training. The technique demonstrates that smaller models can match or exceed larger teacher models' performance when trained with psychological annotations and contextual cues, enabling deployment in resource-constrained environments.
AIBullisharXiv – CS AI · Jun 236/10
🧠Researchers propose VQ4SNN, a hardware-efficient architecture that uses vector quantization to reduce memory requirements for spiking neural networks on FPGAs by 52-61% without sacrificing inference accuracy. This innovation addresses a critical bottleneck in deploying dense SNNs on edge hardware, combining weight-sharing techniques with FPGA-aware memory optimization.
AIBullisharXiv – CS AI · Jun 236/10
🧠Researchers introduce ScalePredictor, a dynamic quantization framework that optimizes Vision Transformer deployment on edge devices by learning instance-aware quantization scales. The method leverages correlations between shallow-layer activation distributions and deeper-layer optimal scales, achieving superior accuracy-efficiency trade-offs compared to existing post-training quantization approaches.
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers introduce ESM (Essential Subspace Merging), a framework that combines multiple task-specific AI models into a single multi-task model by analyzing parameter updates through PCA and projecting them onto essential subspaces. The method reduces task interference while preserving specialized functionality, achieving state-of-the-art performance in model merging without additional training.
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers establish theoretical limits on weight quantization in large language models, identifying 1.58-bit as the minimum precision threshold before expressive collapse occurs. The study demonstrates that model performance degrades polynomially as quantization bits decrease, providing theoretical foundations for optimizing model compression and inference acceleration techniques.
AINeutralarXiv – CS AI · Jun 196/10
🧠Researchers measured the actual linearity of transformer feed-forward network blocks across multiple language models, finding that linearity varies dramatically between adjacent blocks and is learned during training rather than determined by architecture. This discovery enables targeted compression strategies and reveals methodological issues in evaluating transformer models.
🏢 Perplexity
AIBullisharXiv – CS AI · Jun 196/10
🧠Researchers introduce memory optimization techniques for fine-tuning Large Language Models using LoRA on resource-constrained devices, achieving up to 28× peak memory reduction through quantization, efficient checkpointing, and token approximation methods. The work enables private model personalization on consumer hardware without compromising model quality.
🧠 Llama
AIBullisharXiv – CS AI · Jun 196/10
🧠Researchers introduce STORM, a spatial-aware token reduction framework that addresses performance collapse in visual state space models like Mamba when applying token reduction techniques. By maintaining structural integrity and two-dimensional grid topology during compression, STORM achieves significant accuracy recovery, particularly on VMamba with up to 63.3% improvement while operating as a training-free plug-and-play module.
AINeutralarXiv – CS AI · Jun 196/10
🧠Researchers introduce DiverseDistill, a knowledge distillation framework that leverages multiple teachers (foundation models plus domain experts) to more effectively transfer knowledge to compact models. The method recovers 73-114% of the performance gap between teacher and student models while operating with frozen teachers and zero inference overhead.
AINeutralarXiv – CS AI · Jun 196/10
🧠Researchers introduce QC-GAN, a parameter-efficient speech enhancement model combining Quaternion Conformer architecture with MetricGAN training. The framework achieves state-of-the-art speech quality scores while using less than half the parameters of comparable models, with a 35K-parameter variant demonstrating viable ultra-lightweight performance.
