Unverified50% confidenceFactExact time
The researchers constructed a smoothed predictor based on neighboring token-pair counts and evaluated it on GPT-2's full vocabulary of 50,257 tokens.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe researchers conducted downstream evaluation on a GPT-2-scale model with 25 million parameters; the Functionalizer brought significant improvements in code syntactic validity and code character perplexity, while prose text coherence remained largely unchanged.72% similarVerifiedThe GPT series of models are trained on trillion-scale text data through a self-supervised task of predicting the next token.71% similarVerifiedThe BPE (Byte Pair Encoding) algorithm commonly used in the GPT series builds a vocabulary by counting the frequency of adjacent character pairs and merging the highest-frequency character pairs.70% similarUnverifiedGPT-2的BPE在字节级别操作，其merge表包含约5万条规则，最终词表为50,257个token70% similarUnverified类似的极客实验还包括「用 Excel 实现 GPT」、「用正则表达式检测质数」68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001469API
curl https://kongchang.com/api/v1/knowledge/claims/1001469MCP
get_claim(id=1001469)