Few‑shot in‑context learning (ICL) is the ability of a model to infer abstract patterns from input‑output examples provided in its prompt and apply them to new inputs. ICL has been extensively studied in large language models trained for next‑token prediction on human text, and recent work shows it also emerges in autoregressive genomic models.\ \ This paper proposes the Convergent Emergence Hypothesis: when ICL appears across modalities, it shares a common difficulty profile—tasks that benefit from ICL in one modality tend to benefit in others. To test this, the authors build a controlled cross‑modality framework that instantiates the same suite of tasks in six modalities—language, genome, integer sequences, time series, images, and proteins. Results show that paired‑mapping ICL emerges in all six modalities, surpasses controlled baselines, and exhibits correlated per‑task effects across five of them.\ \ These findings support the hypothesis in most modalities, suggesting that ICL may be a domain‑agnostic learning mechanism, though some exceptions remain to be investigated.\ \ Review