End-to-End Test-Time Training: Making Long Context Work Without the Memory Tax
How TTT-E2E achieves constant inference latency regardless of context length by treating long context as a learning problem rather than an architecture problem.
Tagged
#long context
2 articles tagged with #long context.
Why Your LLM Only Uses 10-20% of Its Context Window (And How TITANS Fixes It)
GPT-4's 128K context window? It only uses about 10% effectively. Google's TITANS architecture introduces test-time memory learning that outperforms GPT-4 on long-context tasks with 70x fewer parameters.