Unverified50% confidenceFactExact time
A paper proposes formalizing the test-time scaling of DLMs as the adaptive allocation of inference compute and decoding strategy, and introduces a frequency-aware remasking strategy.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe paper conducts experiments on two types of tasks, code generation and mathematical reasoning, covering five DLM backbone models.72% similarUnverifiedAutoregressive LLMs can already improve reasoning performance through chain-of-thought, RL fine-tuning, multiple sampling, and additional test-time compute.72% similarUnverifiedThe study analyzed the correlation between protocol-specific characteristics (complexity, number of states, clarity of transition logic) and LLM reasoning performance.69% similarUnverifiedFrequency-aware remasking provides a lightweight decoding strategy adjustment that improves reasoning capability without retraining the model.69% similarUnverifiedLLM的每次调用可以被建模为伯努利试验,N次独立调用的通过次数服从二项分布B(N, p)69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154135API
curl https://kongchang.com/api/v1/knowledge/claims/1154135MCP
get_claim(id=1154135)