Overview Papers & code

Inner loop training Outer loop training Truncated gradient unrolling

Long-context reasoning

Reasoning over Needles-in-a-Haystack Multi-Document Question Answering Reasoning over Drops-in-an-Ocean Model Family & Scale Generalization

Robustness analysis

Robusetness to Unseen Test-Time Lengths Robustness to positional biases

Efficiency analysis Future prospects Current Preprint

PERK: Long-Context Reasoning as Test-Time Learning