pzqpzq/Principia

Principia extracts reusable principles, composes those principles into traceable research ideas, and helps researchers inspect why an idea may be worth testing.

What it solves

Principia addresses the challenge of autonomous scientific discovery by bridging the gap between scientific literature and raw data. It prevents the "black box" problem in AI-driven research by ensuring that discovered relationships are not just statistical fits, but are interpretable, grounded in existing scientific principles, and backed by a transparent trail of evidence.

How it works

Principia operates as a research workbench that transforms scientific knowledge and data into a structured object model. It follows a multi-stage discovery workflow:

  1. Inventory & Understanding: The system profiles local datasets and connects variables to research goals and scientific context.
  2. Generation & Evaluation: It proposes candidate mathematical expressions, fits them to development data, and compares alternatives using validation evidence.
  3. Challenge: It applies held-out checks and evidence gates to ensure the relationship is robust before promoting it to a "Rule."
  4. Synthesis: It maps the final Rules alongside the observations and literature Principles that support them.

Technically, it uses a combination of reasoning and vision models (e.g., DeepSeek-V4-Pro) to analyze data and literature, representing knowledge as "Literature Principles," "Meta-Principles," and "Rules" (executable equations with stated scopes).

Who it’s for

It is designed for researchers and scientists across various domains’s (such as physics, seismology, biology, and materials science) who want to automate the discovery of interpretable laws or relationships from their local datasets while maintaining full inspectability of the evidence.

Highlights

  • Autonomous Scientific Discovery (ASD): Automates the path from raw data to interpretable, typeset equations.
  • Evidence-Linked Results: Every discovered Rule is linked to numerical records, calibration results, and validation decisions.
  • Scientific Object Model: Treats scientific claims as revisioned objects with defined boundaries, falsifiers, and provenance rather than simple text.
  • Broad Domain Applicability: Includes a library of 20 public test scenarios spanning fields from astronomy and particle physics to economics and medical imaging.

Related

  • Project
    pzqpzq/LSF_MDia

    MDia is a Python library that formalises LLM intermediate reasoning as reusable “dialect cards” (LSFs) and provides a deterministic eight‑stage pipeline (collect → create → evolve → profile → select → run → validate‑rules → report) for generating, evolving, profiling, routing, and auditing these protocols across heterogeneous models. It supports black‑box LLM APIs, offers several routing plans, includes a 100‑rule bank, and guarantees reproducibility by freezing decisions before test evaluation. The repo includes a toy offline demo that runs without any API keys, extensive documentation, MIT licensing, and a citation to the ICML 2026 poster and arXiv paper.

  • Project
    Xueyang-Song/paper-pilot

    Paper Pilot is a desktop (Electron + React) research assistant that crawls 8+ scholarly databases, downloads PDFs, indexes them with SQLite + FTS5 + vector search, and lets you ask a local or hosted LLM for evidence‑grounded answers. All data stays on your machine; the app provides audit‑ready citation traces and a reproducible workflow for literature review.

  • Project
    ailia-ai/ailia-models

    A collection of 419 pre-trained AI models across various domains like vision, audio, and text, all runnable via a unified CLI with automatic weight downloads.

  • Project
    shibing624/agentica

    Agentica is an open‑source, Apache‑2.0‑licensed framework that lets you run multiple LLM‑powered agents locally. It provides a unified CLI, web UI, and desktop app that share the same history and configuration. Features include async‑first tool execution, multi‑model and multi‑modal support, persistent memory with context compression, a markdown‑based skill system that self‑evolves, safety guardrails, and built‑in collaboration (peer messaging, task/sub‑agent, and delegate processes). Python and TypeScript SDKs let you embed agents in code, and Docker/uv‑tool installers make setup easy.

  • Dispatch
    OpenAI Prism Release

    OpenAI has launched Prism, a free, AI-native LaTeX workspace powered by GPT-5.2 designed to integrate scientific writing, collaboration, and reasoning into a single environment.