All news

8 articles

Min score

Reset

paperarXiv (systems, data, ML)·23 Oct 20264.1
BeNice! Closing Looping's Inference-Memory Gap via KV-Sharing
BeNice shares KV cache across loop iterations in looped transformers to close the inference memory gap.
KV Cache llm-architecture
paperarXiv (systems, data, ML)·22 Oct 20264.1
Mixing Matters: Evidence Position Bias Across Sequence Mixers in Long-Context Question Answering
Evidence position bias differs across attention and other sequence mixers in long-context QA.
llm-architecture long-context
paperarXiv (systems, data, ML)·23 Oct 20264.0
Rethinking Sequence Compression with Test-time Regression
Recasts sequence compression as test-time regression for efficient models.
llm-architecture linear-attention
paperIEEE Xplore·1 Nov 20263.8
BitNet v2: Low-Bit and Sparsified Activations for 1-Bit Large Language Models
BitNet v2 adds low-bit, sparsified activations to 1-bit LLMs to cut inference cost further.
quantization llm-architecture
paperarXiv (systems, data, ML)·22 Oct 20263.5
Cola2: Co-Adaptive Hierarchical Language Modeling with Continuous Semantic Compression and Balanced Discrete Generation
Co-adaptive hierarchical language modeling with continuous semantic compression.
llm-architecture
paperarXiv (systems, data, ML)·22 Oct 20263.4
Delta N-gram: Forward-Only Continual Writing into a Language Model's Native N-gram Memory
Writes continual updates forward-only into an LM native n-gram memory table.
llm-architecture continual-learning
paperGoogle Scholar·25 Oct 20263.4
A stochastic-computing hardware accelerator for rotary position embedding in large language models
Stochastic-computing hardware accelerator for RoPE in LLMs.
ai-accelerators llm-architecture
paperarXiv (systems, data, ML)·22 Oct 20263.2
How Depth Growth and Looping Shape In-Context Learning in Language Models
How depth growth and looping shape in-context learning.
llm-architecture in-context-learning