Unverified50% confidenceBenchmarkExact time
The fine-tuned models achieved consistent improvements across different language pairs, different domains, and both automatic evaluation metrics and human evaluation.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLarge language models are optimized to 'maximize the prediction accuracy of the next token,' which naturally inclines them toward generating 'complete, self-consistent' outputs.78% similarUnverifiedTraditional large language model evaluation centers on the accuracy of single-turn question answering, scoring the output answer against a reference answer after a question is given72% similarUnverifiedEvaluation metrics use each model's own normal prose output as the baseline, and each model's own tokenizer is used to calculate token counts to ensure fairness of comparison.72% similarUnverifiedLarge language models can approximately estimate decision confidence by analyzing token probability distributions, explicitly outputting self-evaluation scores, or observing answer consistency across multiple samplings.72% similarUnverifiedThe direct distribution alignment method can stably improve cross-lingual consistency performance across all model-dataset combinations.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949200API
curl https://kongchang.com/api/v1/knowledge/claims/949200MCP
get_claim(id=949200)