perplexity
A metric commonly used to evaluate language models, measuring how well a probability distribution predicts a sample. A lower perplexity indicates better predictive performance and evokes clarity in language generation tasks.
- A Theoretical Study on Bridging Internal Probability and Self-Consistency for LLM Reasoning
- AlphaDecay: Module-wise Weight Decay for Heavy-Tailed Balancing in LLMs
- Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking
- CLAWS:Creativity detection for LLM-generated solutions using Attention Window of Sections
- Correlation Dimension of Autoregressive Large Language Models
- DenoiseRotator: Enhance Pruning Robustness for LLMs via Importance Concentration
- First Attentions Last: Better Exploiting First Attentions for Efficient Parallel Training
- HBLLM: Wavelet-Enhanced High-Fidelity 1-Bit Quantization for LLMs
- Improving Model Representation and Reducing KV Cache via Skip Connections with First Value Heads
- LOMIA: Label-Only Membership Inference Attacks against Pre-trained Large Vision-Language Models
- RSAVQ: Riemannian Sensitivity-Aware Vector Quantization for Large Language Models
- Routing Mamba: Scaling State Space Models with Mixture-of-Experts Projection
- Tensor Product Attention Is All You Need