Unverified50% confidenceFactExact time
At equivalent parameter scales, diffusion language models typically still lag behind mature autoregressive models in perplexity and downstream task performance.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/9/2026
First Seen
Valid until: 12/8/2026
Sources
Related Entities
Related Claims
UnverifiedSmaller models often show exponential rather than linear degradation in performance on long reasoning chain tasks.77% similarUnverifiedIn long contexts, models are more easily influenced by the most recent few turns of conversation rather than the system prompt set at the beginning, a phenomenon known as instruction-following decay.77% similarUnverifiedPrompt engineering has inherent limitations: the effectiveness of instructions decays over conversation turns (instruction forgetting), and different model versions respond significantly differently to the same instruction.75% similarUnverified量化模型的困惑度可能仅上升0.1-0.5，但在特定下游任务（如代码生成、数学推理）上的表现可能出现显著退化75% similarUnverifiedPrompt engineering remains an empirical technique rather than an exact science; different models respond very differently to the same prompt, and model updates may render old prompts ineffective.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/884174API
curl https://kongchang.com/api/v1/knowledge/claims/884174MCP
get_claim(id=884174)