DLR-Lock locks open-weight LLMs against unauthorized fine-tuning by swapping MLPs for deep low-rank residual networks that inflate backprop memory and complicate optimization, yet preserve original capabilities via module-wise distillation.
hub Tool reference
Pointer sentinel mixture models
Tool reference. 80% of classified Pith citations use this work as a method, library, or software dependency, not as a substantive claim.
12
Pith papers citing it
Method reference
80% of classified citations
hub tools
citation-role summary
dataset 4
other 1
citation-polarity summary
representative citing papers
STS repurposes draft-model attention scores from speculative decoding to build token-and-head-wise sparsity masks, delivering 2.67x speedup at ~90% sparsity on NarrativeQA with negligible accuracy loss.
LAQuant improves long-decoding accuracy on quantized reasoning models like Qwen3-4B by 15pp on AIME25 via layer-wise lookahead loss, achieving 3.42x speedup over FP16.
Pre-trained MoE models exhibit up to 90% intra-expert activation sparsity that enables up to 2.5x faster MoE layer execution when exploited in the vLLM inference system.
PARSE trains a prompt-aware linear router on dense-model outputs to select dynamic SVD ranks, improving accuracy up to 10% at 0.6 compression ratio on LLaMA-7B while delivering 2.5x prefill and 2.4x decode speedups.
Training transformers with KV sparsification during continued pretraining produces representations that admit better post-hoc KV cache compression, improving quality under memory budgets for long-context tasks.
Parcae stabilizes looped LLMs via spectral norm constraints on injection parameters, enabling power-law scaling for training FLOPs and saturating exponential scaling at test time that improves quality over fixed-depth baselines under fixed parameter budgets.
TrainMover achieves ~20s downtime for interruptions in 1024-GPU LLM training via two-phase delta-based communication setup, communication-free sandboxed warmup, and general standby design, projecting 55% reduction in wasted GPU hours.
LLMs show partial and variable perceptual alignment with human touch on textiles, succeeding on samples like silk satin but failing on cotton denim when matching descriptive language to embedding similarity.
ALiBi enables transformers trained on length-1024 sequences to extrapolate to length-2048 with the same perplexity as a sinusoidal model trained on 2048, while training 11% faster and using 11% less memory.
Quant.npu provides a fully static quantization pipeline for on-device LLMs on NPUs by combining rotation matrices, bit-width-aware initialization, two-stage selective optimization, and adaptive mixed precision.
Agentic RAG embeds agents with reflection, planning, tool use, and collaboration into retrieval pipelines to overcome static RAG limitations, and the survey offers a taxonomy by agent count, control, autonomy, and knowledge representation plus applications and open challenges.
citing papers explorer
-
Locking Pretrained Weights via Deep Low-Rank Residual Distillation
DLR-Lock locks open-weight LLMs against unauthorized fine-tuning by swapping MLPs for deep low-rank residual networks that inflate backprop memory and complicate optimization, yet preserve original capabilities via module-wise distillation.
-
STS: Efficient Sparse Attention with Speculative Token Sparsity
STS repurposes draft-model attention scores from speculative decoding to build token-and-head-wise sparsity masks, delivering 2.67x speedup at ~90% sparsity on NarrativeQA with negligible accuracy loss.
-
LAQuant: A Simple Overhead-free Large Reasoning Model Quantization by Layer-wise Lookahead Loss
LAQuant improves long-decoding accuracy on quantized reasoning models like Qwen3-4B by 15pp on AIME25 via layer-wise lookahead loss, achieving 3.42x speedup over FP16.
-
Uncovering Intra-expert Activation Sparsity for Efficient Mixture-of-Expert Model Execution
Pre-trained MoE models exhibit up to 90% intra-expert activation sparsity that enables up to 2.5x faster MoE layer execution when exploited in the vLLM inference system.
-
Different Prompts, Different Ranks: Prompt-aware Dynamic Rank Selection for SVD-based LLM Compression
PARSE trains a prompt-aware linear router on dense-model outputs to select dynamic SVD ranks, improving accuracy up to 10% at 0.6 compression ratio on LLaMA-7B while delivering 2.5x prefill and 2.4x decode speedups.
-
Training Transformers for KV Cache Compressibility
Training transformers with KV sparsification during continued pretraining produces representations that admit better post-hoc KV cache compression, improving quality under memory budgets for long-context tasks.
-
Parcae: Scaling Laws For Stable Looped Language Models
Parcae stabilizes looped LLMs via spectral norm constraints on injection parameters, enabling power-law scaling for training FLOPs and saturating exponential scaling at test time that improves quality over fixed-depth baselines under fixed parameter budgets.
-
TrainMover: An Interruption-Resilient Runtime for ML Training
TrainMover achieves ~20s downtime for interruptions in 1024-GPU LLM training via two-phase delta-based communication setup, communication-free sandboxed warmup, and general standby design, projecting 55% reduction in wasted GPU hours.
-
TouchAI: Exploring human-AI perceptual alignment in touch through language model representations
LLMs show partial and variable perceptual alignment with human touch on textiles, succeeding on samples like silk satin but failing on cotton denim when matching descriptive language to embedding similarity.
-
Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
ALiBi enables transformers trained on length-1024 sequences to extrapolate to length-2048 with the same perplexity as a sinusoidal model trained on 2048, while training 11% faster and using 11% less memory.
-
Quant.npu: Enabling Efficient Mobile NPU Inference for on-device LLMs via Fully Static Quantization
Quant.npu provides a fully static quantization pipeline for on-device LLMs on NPUs by combining rotation matrices, bit-width-aware initialization, two-stage selective optimization, and adaptive mixed precision.
-
Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG
Agentic RAG embeds agents with reflection, planning, tool use, and collaboration into retrieval pipelines to overcome static RAG limitations, and the survey offers a taxonomy by agent count, control, autonomy, and knowledge representation plus applications and open challenges.