DuoAttention identifies retrieval heads requiring full KV cache and streaming heads using constant-length cache to reduce memory and latency in long-context LLM inference.
Measuring massive multitask language understanding
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
SDM sparsifies the Gated DeltaNet update rule to enable 1000x larger recurrent memory states at iso-FLOP, improving long-context recall and short-context reasoning over GDN and matching full attention at 8B scale.
Gradient-informed placement of LoRA parameters recovers full performance under GRPO while random placement does not, due to differences in gradient rank and stability across training regimes.
ReasoningGuard is an inference-time method that uses attention mechanisms to inject safety aha moments and scaling sampling to defend large reasoning models against jailbreak attacks.
TAH-Quant introduces tile-wise adaptive Hadamard quantization for activations in pipeline parallelism, achieving 3-4 bit compression with up to 4.3x throughput speedup and O(1/sqrt(T)) convergence matching SGD.
Atlas reaches over 42% accuracy on Natural Questions with only 64 examples, outperforming a 540B-parameter model by 3% with 50x fewer parameters.
LoRA-FA freezes LoRA's A matrix and trains only B with gradient corrections to approximate full fine-tuning gradients more closely.
A survey deriving a unified policy gradient framework for LLM post-training methods and providing technical comparisons of PPO, GRPO, DPO variants.
citing papers explorer
-
DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads
DuoAttention identifies retrieval heads requiring full KV cache and streaming heads using constant-length cache to reduce memory and latency in long-context LLM inference.
-
Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity
SDM sparsifies the Gated DeltaNet update rule to enable 1000x larger recurrent memory states at iso-FLOP, improving long-context recall and short-context reasoning over GDN and matching full attention at 8B scale.
-
Not How Many, But Which: Parameter Placement in Low-Rank Adaptation
Gradient-informed placement of LoRA parameters recovers full performance under GRPO while random placement does not, due to differences in gradient rank and stability across training regimes.
-
ReasoningGuard: Safeguarding Large Reasoning Models with Inference-time Safety Aha Moments
ReasoningGuard is an inference-time method that uses attention mechanisms to inject safety aha moments and scaling sampling to defend large reasoning models against jailbreak attacks.
-
TAH-QUANT: Effective Activation Quantization in Pipeline Parallelism over Slow Network
TAH-Quant introduces tile-wise adaptive Hadamard quantization for activations in pipeline parallelism, achieving 3-4 bit compression with up to 4.3x throughput speedup and O(1/sqrt(T)) convergence matching SGD.
-
Atlas: Few-shot Learning with Retrieval Augmented Language Models
Atlas reaches over 42% accuracy on Natural Questions with only 64 examples, outperforming a 540B-parameter model by 3% with 50x fewer parameters.
-
LoRA-FA: Efficient and Effective Low Rank Representation Fine-tuning
LoRA-FA freezes LoRA's A matrix and trains only B with gradient corrections to approximate full fine-tuning gradients more closely.
-
Reinforcement Learning for LLM Post-Training: A Survey
A survey deriving a unified policy gradient framework for LLM post-training methods and providing technical comparisons of PPO, GRPO, DPO variants.