OpenRFM CLAUDE.md

OpenRFM CLAUDE.md is an instructions file for coding agents from T-Lab/OpenRFM. It costs 4,694 tokens per session, scanned A, original, MIT.

A repository-specific instruction file for OpenRFM, a reproduction of a relational foundation model used with supply-chain data benchmarks. It documents architecture, training stability, and time-related correctness rules.

In plain words
What is it for?
Use it when working on the OpenRFM implementation to follow its training and evaluation workflow, preserve module architecture, and run the documented syntax checks.
Why use it?
It prevents coding changes that would invalidate the intended experiment, such as fine-tuning separately for each benchmark task instead of evaluating with in-context learning.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/t-lab/openrfm/claude-md
Clone the repo
git clone --depth 1 https://github.com/T-Lab/OpenRFM

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for OpenRFM CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/t-lab/openrfm/claude-md.svg)](https://agentmods.dev/instructions/t-lab/openrfm/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/t-lab/openrfm/claude-md"><img src="https://agentmods.dev/badge/instructions/t-lab/openrfm/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 4,694 This file is loaded in full into every session.
When invoked 4,694 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.04694 $0.04694
Opus 5 $0.02347 $0.02347
Sonnet 5 $0.00939 $0.00939
Haiku 4.5 $0.00469 $0.00469

Measured 4d ago against content hash 409ed432af77, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

OpenRFM CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 204 lines

How it starts

The opening of the file, as written. The whole thing — 204 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Overview

KumoRFM-2 reproduction: a relational foundation model pre-trained on synthetic SCM tasks, then evaluated in-context (no fine-tuning) on RelBench benchmarks. Source paper: technical_report/KumoRFM-technical-report.pdf.

Package: implementation/kumorfm_repro/. Work from implementation/.

cd /home/user/work/KumoRFM/implementation
python -m compileall kumorfm_repro   # fast syntax check

Critical Architectural Constraints

The paper's approach is pre-train on synthetic data once, then evaluate on RelBench via in-context learning (ICL). Do NOT fine-tune on each RelBench task separately.

The benchmark_adapter train-model command does per-task fine-tuning. Use it only for smoke testing the file-backed adapter. For reproduction evidence, use:

  1. kumorfm_repro.train — synthetic pre-training (produces a checkpoint)
  2. kumorfm_repro.icl_eval or kumorfm_repro.icl_suite — ICL evaluation on RelBench tasks using that checkpoint

Training Stability Rules

After extensive experimentation, these are the only known stable configurations:

Config AMP Context Model Stability
Baseline + AMP yes ≤24 AttnBlock/TableEncoder Stable (99.9% step success)
Improved + no AMP no ≤16 ImprovedAttnBlock/ImprovedTableEncoder Stable (untested at scale)
Improved + AMP yes ≤16 ImprovedAttnBlock + safe pooling + FP32 MHA + typed graph attention 120-step d_model=64 typed-graph smoke stable with low AMP scale; not long-run-proven
Baseline + AMP yes ≥32 AttnBlock NaN in all steps
d_model ≥ 384 any any any NaN

The Improved model (RoPE, attention pooling, typed pairwise graph attention) has richer representations and now has an AMP-safe short-run path after fixing all-masked attention pooling, running MHA in FP32 under autocast, and lowering the default GradScaler initial scale. It has passed a 120-step, context-16, d_model-64 two-GPU AMP smoke with zero skipped steps and a 500-step version with saved best checkpoint. The 500-step checkpoint shows context-target sensitivity on F1 ICL ablations, but has not yet beaten the root-only average, so the Baseline model with AMP remains the only long-run-proven config.

Read the full file on GitHub · 204 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 204 lines · 4,694 tokens per session scan A 409ed432af77

Subscribe to this mod's changes

OpenRFM CLAUDE.md is an instructions file published in the GitHub repository T-Lab/OpenRFM (22 stars, last pushed 3mo ago), licensed MIT. It adds 4,694 tokens to every session, about $0.0235 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other instructions, from other repositories

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 6,785 tokens

spec-kit AGENTS.md

AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.

github/spec-kit · 7,104 tokens

codex AGENTS.md

AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.

openai/codex · 5,182 tokens

langchain AGENTS.md

AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.

langchain-ai/langchain · 4,345 tokens

vscode oss-third-party-notices.instructions.md

Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).

microsoft/vscode · 5,001 tokens

next.js AGENTS.md

Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.

vercel/next.js · 7,296 tokens