![[2602.21317] Shared Nature, Unique Nurture: PRISM for Pluralistic Reasoning via In-context Structure Modeling](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.21317] Shared Nature, Unique Nurture: PRISM for Pluralistic Reasoning via In-context Structure Modeling
Summary
The paper presents PRISM, a model-agnostic system designed to enhance large language models (LLMs) by fostering pluralistic reasoning through individualized epistemic trajectories, achieving significant improvements in creativity and diagnostic accuracy.
Why It Matters
This research addresses the limitations of current LLMs, which often converge towards a singular perspective, hindering creativity and scientific discovery. By introducing PRISM, the authors propose a new paradigm that encourages diverse cognitive approaches, potentially transforming AI's role in creative and diagnostic tasks.
Key Takeaways
- PRISM enhances LLMs by integrating individualized epistemic trajectories.
- The model achieves state-of-the-art novelty on creativity benchmarks.
- PRISM demonstrates improved diagnostic capabilities in rare diseases.
- The approach encourages a diverse ecosystem of cognitive perspectives.
- This research proposes a shift from monolithic AI to pluralistic reasoning.
Computer Science > Machine Learning arXiv:2602.21317 (cs) [Submitted on 24 Feb 2026] Title:Shared Nature, Unique Nurture: PRISM for Pluralistic Reasoning via In-context Structure Modeling Authors:Guancheng Tu, Shiyang Zhang, Tianyu Zhang, Yi Zhang, Diji Yang View a PDF of the paper titled Shared Nature, Unique Nurture: PRISM for Pluralistic Reasoning via In-context Structure Modeling, by Guancheng Tu and 4 other authors View PDF HTML (experimental) Abstract:Large Language Models (LLMs) are converging towards a singular Artificial Hivemind, where shared Nature (pre-training priors) result in a profound collapse of distributional diversity, limiting the distinct perspectives necessary for creative exploration and scientific discovery. To address this, we propose to equip models with inference-time Nurture (individualized epistemic trajectories) using Epistemic Evolution paradigm, progressing through explore, internalize, and express. We instantiate this via PRISM (Pluralistic Reasoning via In-context Structure Modeling), a model-agnostic system that augments LLM with dynamic On-the-fly Epistemic Graphs. On three creativity benchmarks, PRISM achieves state-of-the-art novelty and significantly expands distributional diversity. Moreover, we evaluate the real-world utility via a challenging rare-disease diagnosis benchmark. Results demonstrate that PRISM successfully uncovers correct long-tail diagnoses that standard LLM miss, confirming that its divergence stems from meaningful...