---
title: "Research"
description: "Research from Nace.AI on training trustworthy models capable of executing tasks reliably while continuously self-improving."
canonical: "https://nace.ai/research"
updated: 2026-07-22
---

# Research

> Research from Nace.AI on training trustworthy models capable of executing tasks reliably while continuously self-improving.

## We conduct focused research in Applied Meta-Learning.

Many challenges in AI remain beyond the reach of scale alone.

## Featured

- [Scaling Laws for Hypernetwork-Based Knowledge Injection](https://nace.ai/research/scaling-law-for-hypernetwork): New scaling laws for train-time knowledge injection: hypernetworks beat LoRA and full fine-tuning on out-of-distribution generalization, and the gap grows with target model size. (2026-07-22)
- [Enterprise Policy Injection with Metamodels](https://nace.ai/research/enterprise-policy-injection-with-metamodels): Adapting model behavior at inference time without retraining, large context windows, or touching the primary model's weights. (2026-05-01)
- [Nace Meta Agent](https://nace.ai/research/nacemetaagent): NEMA: meta-learning architectures that generate specialized, task-specific agents automatically. (2026-01-27)
- [NAVI Console: Policy-driven Safeguards Comparison](https://nace.ai/research/policy-driven-safeguards): Benchmarking Nace's Safety Model against AWS Bedrock, Azure, NeMo Guardrails, and prompt-based approaches. (2025-01-19)

## All publications

- [Scaling Laws for Hypernetwork-Based Knowledge Injection](https://nace.ai/research/scaling-law-for-hypernetwork): New scaling laws for train-time knowledge injection: hypernetworks beat LoRA and full fine-tuning on out-of-distribution generalization, and the gap grows with target model size. (2026-07-22)
- [Enterprise Policy Injection with Metamodels](https://nace.ai/research/enterprise-policy-injection-with-metamodels): Adapting model behavior at inference time without retraining, large context windows, or touching the primary model's weights. (2026-05-01)
- [Nace Meta Agent](https://nace.ai/research/nacemetaagent): NEMA: meta-learning architectures that generate specialized, task-specific agents automatically. (2026-01-27)
- [Memory Augmentation and Editing Techniques in LLMs](https://nace.ai/research/memory-augmentation-and-editing-techniques-in-large-language-models): Memory augmentation and model editing for handling outdated and domain-specific knowledge. (2025-03-18)
- [NAVI Console: Policy-driven Safeguards Comparison](https://nace.ai/research/policy-driven-safeguards): Benchmarking Nace's Safety Model against AWS Bedrock, Azure, NeMo Guardrails, and prompt-based approaches. (2025-01-19)
- [GeLoRA: Efficient Fine-Tuning of Large Language Models](https://nace.ai/research/gelora): A theoretical framework connecting the geometry of data representations to training dynamics. (2024-12-10)
- [Hypernetworks for Specialized Instructions](https://nace.ai/research/hypernetworks-for-specialized-instructions): Efficient fine-tuning of LLMs through task-specific parameter generation. (2024-12-07)
- [Meta-Learning through Hypernetworks](https://nace.ai/research/meta-learning-through-hypernetworks): Improving task adaptability in large language models. (2024-12-07)
- [Memory Augmentation for Document-heavy Enterprise Use Cases](https://nace.ai/research/memory-augmentation-for-document-heavy-enterprises): A modular RAG pipeline tailored to enterprise needs: over 90% F1 on verification. (2024-11-05)
- [Knowledge Graph Synergy with LLMs](https://nace.ai/research/knowledge-graph-utilization): How structured, verifiable knowledge complements LLMs where they fall short. No ML background required. (2024-11-01)
- [Progressing Towards Meta Models and What Lies Beyond](https://nace.ai/research/progressing-towards-meta-models-and-what-lies-beyond): Tailored, meta-learned Small Language Models for rapid, efficient adaptation. (2024-10-24)
- [Understand LLMs From the Perspective of Ranking](https://nace.ai/research/understand-llms-from-the-perspective-of-ranking): Most post-training performance gains are reranking: aligning the proposal distribution with the task. (2024-10-12)

## For agents

- [Drex API guide](https://nace.ai/drex.md): the decision model: request shape, question types, benchmarks and deployment, as markdown
- [Decision endpoint](https://console.nace.ai/v1/systemone): POST with model "drex-latest", a state string and typed questions (choice, noul, score); https://console.nace.ai/v1/models lists checkpoints
- [Document endpoints](https://console.nace.ai/v1/documents/{parse|split|classify|extract|ground}): POST a source (a public URL, an uploaded workspace file or a finished parse job) and poll GET https://console.nace.ai/v1/documents/jobs/{id}; the same key and credit as decisions
- [Console](https://console.nace.ai): API keys, the playground, usage and billing; send `Authorization: Bearer $NACE_API_KEY` on every request
- [Console docs](https://console.nace.ai/docs): quickstart, question types, the document routes, errors, limits and pricing
- [Nace Document Intelligence](https://nace.ai/ndi.md): the Perception series: parse, split, classify, extract and ground any file type with a location on every value
- [Developers](https://nace.ai/developers.md): Console: what it holds, a scripted agent session, the ways in and the four deployment boundaries, from our managed cloud to air-gapped
- [Site index](https://nace.ai/llms.txt): every page on nace.ai; append `.md` to any page URL or send `Accept: text/markdown` for the markdown version