ByteDance Seed says it found cause of DeepSeek's long-context performance drift
AI summary
According to a DoNews report, ByteDance's Seed team says it has identified the cause of DeepSeek models "acting up" (producing erratic outputs), pointing to possible performance drift in long-context scenarios. The report carries headline-level information only, with no model versions, test methods or detailed conclusions given, and no public response from ByteDance or DeepSeek.
Why it matters: If long contexts really can trigger performance drift, the finding would be relevant to anyone building on long-context models, though it still needs a paper or replication to confirm.
How we got here
- Google's Gemini agent can route work between Gemini and Claude虎嗅 AI · ByteDance
- ByteDance Seed ties long-context swings to chunked KV cache 'phase sensitivity'AIbase AI新闻 · ByteDance
- China's tech giants raise ~$56B in debt and equity to pre-buy AI capacity36氪 人工智能 · ByteDance
- ByteDance Seed paper ties DeepSeek long-context swings to KV-cache phase sensitivityDoNews · ByteDance Seed
- Doubao app adds utility bill payment via voice or textDoNews · ByteDance
- Study: Only 3.6% of 857 Chinese frontier AI releases disclosed safety resultsSemiAnalysis · ByteDance