paper

CLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token Inference

paperactiveprovisionalclp-collocation-length-prediction-for-zero-loss-adaptive-multi-token-inference-9538e306·1 events·first seen 7d ago

Aliases: CLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token Inference

Co-occurring entities

Qwen2.5 Alibaba CLP (Collocation-Length Predictor)Backbone-as-Architect

More like this (12)

CLP (Collocation-Length Predictor)Multi-Token Prediction (MTP)Continual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMs Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It Beyond Uniform Tokens: Adaptive Compression for Time Series Language Models PC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-Training Language Modeling Loss When Does Mixing Help? Analyzing Query Embedding Interpolation in Multilingual Dense Retrieval Alternating Token-Weighted Unlearning Planning-aligned Token Compression for Long-Context Autonomous Driving FlashbackCL: Mitigating Temporal Forgetting in Federated Learning ExpRL: Exploratory RL for LLM Mid-Training

Recent events (1)

5arXiv · cs.AI·7d ago·source ↗

CLP: Lightweight collocation-length predictor achieves zero-loss multi-token inference speedup

Researchers propose CLP (Collocation-Length Predictor), a span-level decision layer for accelerating LLM inference via multi-token prediction without quality degradation. The key insight is 'Backbone-as-Architect': the backbone LM head always generates the first token while MTP heads handle only subsequent tokens, eliminating head-backbone competition that causes repetitive outputs in prior methods. CLP uses a single linear layer (~4.6K–7.7K parameters) versus 1M-parameter gate networks in prior work, achieving 1.14x–1.29x speedup on Qwen2.5 models with near-zero repetition ratio. The paper also establishes that shorter prediction horizons improve MTP head accuracy on larger models, offering a scaling-aware design principle.

Inference Economics Qwen2.5 Alibaba CLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token Inference +2 more