A user on Reddit reported that Anthropic's Opus 5 model experienced a decline in performance over time, producing convoluted and overly verbose responses. The user found that deleting the model's memory significantly improved its output, restoring it to a more understandable and useful state similar to its initial performance. This suggests that the model's memory may have accumulated unintended instructions that negatively impacted its responses. AI
IMPACT Suggests potential issues with long-term memory accumulation in LLMs, impacting user experience and requiring workarounds.
RANK_REASON User report on model performance degradation and a workaround.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
