The Ratchet Effect in Silico through Interaction-Driven Cumulative Intelligence in Large Language Models

TL;DR

This paper introduces POLIS, a social interaction framework for large language models that enables cumulative knowledge growth through peer verification and internalization, mimicking human cultural evolution.

Contribution

It demonstrates that structured social interactions among models can significantly improve reasoning performance and knowledge retention, independent of parameter size.

Findings

Models of 1-4 billion parameters improved by 8.8-18.9 points on reasoning benchmarks.

Peer verification is identified as the key mechanism for knowledge ratcheting.

Internalization sustains knowledge accumulation across multiple interaction rounds.

Abstract

Human intelligence scales through cumulative cultural evolution (CCE), a ratchet process in which innovations are retained against entropic drift. Large language model training, by contrast, still depends primarily on static corpora and parameter growth, leaving little room for endogenous accumulation through interaction. We present POLIS (Population Orchestrated Learning and Inference Society), a framework in which heterogeneous agents generate solutions, verify one another's outputs, retain validated artifacts in shared cultural memory, and internalize them through parameter updates. On mathematical reasoning benchmarks, populations of 1--4B-parameter models achieved average gains of 8.8--18.9 points over base models and narrowed the gap to 70B+ monoliths. Mechanistic ablations identify peer verification as the main ratchet operator and show that internalization sustains accumulation across rounds, providing computational evidence that epistemic vigilance organizes durable knowledge growth. These results position structured social interaction as a scaling lever orthogonal to parameter count.