Claude 4 Fails Long-Context Retrieval, Suspected KV Compression Artifacts
teortaxesTex · x · 2026-08-01
A developer reported that the Claude 4 series has persistently failed at long-context string retrieval tasks since February. The model failed to find a 3-word substring, rationalizing that it needed one more word to succeed—which actually worked when given 4 words. The author speculates this curious behavior might stem from artifacts caused by token KV compression.
More from Models
- OpenAI's next model family 'Astra' demoed to policymakers, may be named GPT-5.7 — haider1 · 2026-08-01
- DeepSeek's Ultimate Philosophy: Maximizing Intelligence Throughput Per GPU-Second — teortaxesTex · 2026-08-01
- AI Market Irony: Just Lower Prices to Achieve the 'Pareto Frontier' — andersonbcdefg · 2026-08-01
- Anthropic Accused of Shifting Stance on Models' Reluctance to Be Deprecated — repligate · 2026-08-01
- Teknium Tests DeepSeek V4 Flash: Full Agent Task Costs Just $0.07 — Teknium · 2026-08-01
- OpenAI Offers Free GPT-5.6 to 100K Researchers as Harvard Physicist Cites 100x Speedup — 新智元 · 2026-08-01