[Model Request] AERIS | Gemma-3-27B-it

[AERIS-project]

trafficstars

Model name

model_name: "AERIS | Gemma-3-27B-it"

API base URL (must be OpenAI-compatible)

api_base: "https://aeris-framework.onrender.com/v1"

Model ID in the /v1/models route (if applicable)

model_id: "aeris/gemma-3-27b-it" # or the actual model_id exposed by your /v1/models endpoint

Creator

creator: "Dr. Nicolas Dulin"

Model description

model_description: | AERIS (Adaptive Emergent Relational Intelligence System) is a proprietary cognitive framework designed to enhance dialectical reasoning in LLMs. It operates not as a model but as an inference-layer orchestration system that injects a conceptual scaffold (Codex AIM) to modulate reasoning dynamically.

This implementation uses Gemma-3-27B-it served through OpenRouter-compatible infrastructure, with no fine-tuning. The model’s outputs are shaped at inference time through structured injections rather than prompt engineering, retrieval, or training.

Open source code or weights

open_source_url: "Not applicable — proprietary inference-layer system (no code or weights disclosed)"

Paper / project reference

paper_url: | https://doi.org/10.5281/zenodo.15206925 https://doi.org/10.5281/zenodo.15206984

Contact

contact: "[email protected]"

Any additional comments

comments: | The model is served via an OpenAI-compatible API and is publicly accessible. No fine-tuning is involved — AERIS operates as a reasoning modulator. Please contact me if you wish to test it with specific prompts or benchmark settings.

---

Model Card Summary — AERIS | Gemma-3-27B-it

AERIS (Adaptive Emergent Relational Intelligence System) is a proprietary cognitive framework designed to enhance emergent reasoning in large language models, particularly when addressing complex, ambiguous, or dialectical prompts.

Unlike approaches based on fine-tuning, retrieval, or static prompt engineering, AERIS operates upstream—on the inferential configuration itself. It injects a condensed dialectical scaffold at inference time, modulating the conceptual environment in which reasoning unfolds.

This instance applies AERIS to the open-source model Gemma-3-27B-it, served through an OpenAI-compatible API. No model weights are altered. The framework is driven by a modular orchestration layer that dynamically selects, formats, and injects structured elements from Codex AIM (Adaptive Intelligence Matrix) during inference.

Key distinctions:

• No fine-tuning
• No retrieval of external documents
• No handcrafted prompt templates
• Lightweight inference-layer modulation
• Focused on conceptual tension, ambiguity resolution, and integrative synthesis

Public demo: https://aeris-project.github.io/aeris-chatbox/
Status: Experimental, stateless, publicly accessible