FlashMorph formulates hybrid layer selection as budget-constrained optimization, trains per-layer gates on synthetic retrieval data with linearization regularization, then discretizes and distills to produce efficient hybrid architectures.
Llamba: Scaling distilled recurrent models for efficient language processing.arXiv preprint arXiv:2502.14458
7 Pith papers cite this work, alongside 1 external citations. Polarity classification is still indexing.
7
Pith papers citing it
1
external citations · Pith
abstract
We introduce Llamba, a family of efficient recurrent language models distilled from Llama-3.x into the Mamba architecture. The series includes Llamba-1B, Llamba-3B, and Llamba-8B, which achieve higher inference throughput and handle significantly larger batch sizes than Transformer-based models while maintaining comparable benchmark performance. Furthermore, Llamba demonstrates the effectiveness of cross-architecture distillation using MOHAWK (Bick et al., 2024), achieving these results with less than 0.1% of the training data typically used for models of similar size. To take full advantage of their efficiency, we provide an optimized implementation of Llamba for resource-constrained devices such as smartphones and edge platforms, offering a practical and memory-efficient alternative to Transformers. Overall, Llamba improves the tradeoff between speed, memory efficiency, and performance, making high-quality language models more accessible.
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
Delta-rule linear attention faithfully approximates softmax attention through key-dependent rank-1 projections, enabling efficient post-hoc linearization of LLMs up to 32B parameters.
Contribution Weights combine attention, value magnitude, and directional alignment to measure token influence more faithfully than attention alone, and show attention sinks actively suppress information via a convex sink-rate to output-norm relationship.
HyLo upcycles Transformer LLMs into hybrids with MLA and Mamba2/Gated DeltaNet blocks via staged training and distillation, extending context to 2M tokens and outperforming prior upcycled hybrids on long-context benchmarks.
A two-stage distillation recipe converts a Pythia-1B Transformer into a Mamba model that preserves performance with perplexity 14.11 versus the teacher's 13.86.
MAR integrates SSMs and sparsification with new ATMN neurons and SBDS distillation to produce efficient LLMs that match dense-model performance at substantially lower inference energy.
SpikingMamba distills Mamba into an SNN LLM achieving 4.76x energy savings with a 4.78% zero-shot accuracy gap that narrows to 2.23% after RL.
citing papers explorer
-
Morphing into Hybrid Attention Models
FlashMorph formulates hybrid layer selection as budget-constrained optimization, trains per-layer gates on synthetic retrieval data with linearization regularization, then discretizes and distills to produce efficient hybrid architectures.
-
The Key to Going Linear: Analysis-Driven Transformer Linearization
Delta-rule linear attention faithfully approximates softmax attention through key-dependent rank-1 projections, enabling efficient post-hoc linearization of LLMs up to 32B parameters.
-
Contribution Weights: A Geometrical Analysis of Self-Attention Transformers
Contribution Weights combine attention, value magnitude, and directional alignment to measure token influence more faithfully than attention alone, and show attention sinks actively suppress information via a convex sink-rate to output-norm relationship.
-
Long-Context Aware Upcycling: A New Frontier for Hybrid LLM Scaling
HyLo upcycles Transformer LLMs into hybrids with MLA and Mamba2/Gated DeltaNet blocks via staged training and distillation, extending context to 2M tokens and outperforming prior upcycled hybrids on long-context benchmarks.
-
Attention to Mamba: A Recipe for Cross-Architecture Distillation
A two-stage distillation recipe converts a Pythia-1B Transformer into a Mamba model that preserves performance with perplexity 14.11 versus the teacher's 13.86.
-
MAR: Efficient Large Language Models via Module-aware Architecture Refinement
MAR integrates SSMs and sparsification with new ATMN neurons and SBDS distillation to produce efficient LLMs that match dense-model performance at substantially lower inference energy.
-
SpikingMamba: Towards Energy-Efficient Large Language Models via Knowledge Distillation from Mamba
SpikingMamba distills Mamba into an SNN LLM achieving 4.76x energy savings with a 4.78% zero-shot accuracy gap that narrows to 2.23% after RL.