Reading Between the Dots: Decoding Hidden Computation across Filler Tokens
Quick summary
arXiv:2607.03502v2 Announce Type: replace-cross Abstract: Frontier LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT). This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning. But hidden from the output is not the same as hidden from us. On four task families (fact retrieval, parallel numeric composition, string manipulation, and in-context computation), two open-weights frontier models (DeepSeek V3, Kimi K2) comput
Key takeaways
- arXiv:2607.03502v2 Announce Type: replace-cross Abstract: Frontier LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT).
- This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning.
- But hidden from the output is not the same as hidden from us.
Why it matters
“Reading Between the Dots: Decoding Hidden Computation across Filler Tokens” should be evaluated beyond branding and benchmark scores. Its practical importance will emerge in task accuracy, latency, unit cost, safety and integration with real workflows.
Member comments