File size: 3,478 Bytes
def3e11 2cb3b06 def3e11 2cb3b06 def3e11 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 |
---
license: apache-2.0
---
https://huggingface.co/datasets/neomonde-lab/law-e-framework/resolve/main/README.md
# Law E Framework – Thermodynamic Governance for AI Reliability
**Short description**
Law E is an operational framework that treats modern AI systems as thermodynamic information processes.
It introduces a native governance layer that observes the “energy cost” and coherence of model outputs, and uses this signal to regulate hallucinations and unstable behaviors.
This repository hosts the **initial technical report** describing the framework, its main equations and the first proof-of-concept design.
📄 **PDF**: [`Law_E_Framework.pdf`](./Law_E_Framework.pdf)
---
## Why Law E?
- Large language models are powerful but prone to **hallucinations**.
- Current guardrails are mostly **symbolic or heuristic**.
- Law E proposes a **physics-inspired governance layer**:
- monitors useless energy dissipation ΔE
- tracks global organization / stability
- regulates inference when the system drifts
The goal is to move toward **self-regulated, energy-aware AI systems**.
---
## Current status
- Conceptual framework and equations defined.
- First regulator–selector POC in development.
- Next steps:
- standardized hallucination evaluation (TruthfulQA, etc.)
- CPU/energy proxy metrics
- public demonstrator for selected models.
---
## Contact & Collaboration
Created by **Sébastien Favre-Lecca (Neomundi Lab)**
- Website: https://neomonde.io
- Twitter:
If you are working on AI safety, energy-aware AI, or robotics and want to collaborate on Law E evaluation or implementation, feel free to reach out.
------------------------------------------------
---
## Experimental Reports (Zenodo mirror)
This repository also hosts a set of experimental research reports archived on Zenodo,
documenting empirical observations and audit-level traces related to information–energy
dynamics and endogenous regulation in artificial systems (Loi E).
### ### Available experimental reports
- **Information–Energy Dynamics Simulator (Loi E) — Experimental Observations**
Systematic exploration of information–energy regimes in a minimal simulator, including
non-uniform dissipation, distributed asymmetry, recoverability windows, and irreversibility thresholds.
📄 PDF: `Information-energy dynamics simulator Loi E — Experimental observations.pdf`
🔗 Zenodo DOI: https://doi.org/10.5281/zenodo.18483571
- **Regulator–Selector (Reg-Sel): Measured effects of endogenous regulation in large language models**
Experimental evaluation of a minimal inference-time regulation mechanism applied to LLMs,
without modifying model weights, prompts, or using external supervision.
📄 PDF: `Measured effects of endogenous regulation in large language models.pdf`
- **Filtre E — Audit trace example: information–energy regulation**
Audit-level illustration of explicit information–energy thresholding and regulated versus
non-regulated response regimes.
📄 PDF: `Filtre E — audit trace example information-energy regulation.pdf`
These documents form a coherent experimental foundation complementing the Law E framework.
Interpretative discussions (e.g. negentropy, biological analogy, governance implications) are
intentionally addressed in separate publications.
Further experimental reports and technical notes will be added to this repository as the Law E research progresses.
|