# Instruction-Level Weight Shaping: A Framework for Self-Improving AI Agents

arXiv: https://arxiv.org/abs/2509.00251 · HTML: https://mlsanity.com/paper/2509.00251/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Agentic RAG, agents and memory, one of 15 directions in Retrieval-augmented generation.
- Builds on 6 of its 8 references that are in the map.
- First posted on arXiv 29 Aug 2025.

## Placed in

- Retrieval-augmented generation / [Agentic RAG, agents and memory](https://mlsanity.com/rag/agentic-rag-agents-and-memory/)
- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)
- Synthetic data and self-improvement for LLMs / [Code and instruction data synthesis](https://mlsanity.com/synth/code-and-instruction-data-synthesis/)

## Builds on (most cited first, 6 of 6)

- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [Demonstrate-Search-Predict: Composing retrieval and language models for knowledge-intensive NLP](https://mlsanity.com/paper/2212.14024/) (2022)
- [The Lipschitz Constant of Self-Attention](https://mlsanity.com/paper/2006.04710/) (2020)
- [A Survey on Retrieval-Augmented Text Generation for Large Language Models](https://mlsanity.com/paper/2404.10981/) (2024)
- [Every Model Learned by Gradient Descent Is Approximately a Kernel Machine](https://mlsanity.com/paper/2012.00152/) (2020)
- [Learning without training: The implicit dynamics of in-context learning](https://mlsanity.com/paper/2507.16003/) (2025)

## Built on by (most cited first, 2 of 2)

- [From Failed Trajectories to Reliable LLM Agents: Diagnosing and Repairing Harness Flaws](https://mlsanity.com/paper/2606.06324/) (2026)
- [Diagnosis-Driven Automatic Repair for Agentic Workflow via Symbolic Inference](https://mlsanity.com/paper/2607.02882/) (2026)

## Abstract (arXiv)

Large language models (LLMs) are fluent but largely static after pre-training; new or shifting knowledge is typically added with retrieval-augmented generation (RAG) or fine-tuning. RAG raises latency and engineering overhead and often fails to integrate facts; prompt engineering is brittle and can conflict with prior knowledge; fine-tuning is costly and risks catastrophic forgetting. We propose Instruction-Level Weight Shaping (ILWS): curated system instructions act as external, auditable pseudo-parameters updated after each session via reflection and user feedback. A Reflection Engine inspects conversation traces, diagnoses reasoning successes and failures, and proposes typed deltas Delta K=(Delta S,Delta U,Delta T) over instructions, user preferences, and tools. Deltas are version-controlled, evaluated with a sliding window of 1-5 star ratings, auto-repaired on first failure, and rolled back on repeated failure. When an edit budget crosses a threshold, the agent compiles a rating-weighted synthetic set and distills matured instruction-space gains into parameters, converting prompt-space improvements into weight-space without downtime. ILWS makes explicit the low-rank shaping induced by context in transformer blocks, preserves governance, and removes per-call retrieval. In enterprise support it increased throughput 2.4-5.0x and cut audited hallucinations by about 80% versus a frozen baseline. In an Adobe Commerce Cloud proof of concept "L0 Support", it achieved 4-5x more tickets per hour and about 80% lower time per ticket, with autonomous instruction updates and optional tool synthesis. Because ILWS operates at the instruction layer until controlled distillation, it generalizes to dynamic domains (legal, medical, engineering) requiring adaptive reasoning, tool creation, and low-latency deployment.