Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 14, 2026Bogota

Evaluación holística y multi-metodológica de propuestas de seguridad en IA

Mateo Acosta

Submitted to AI Incident Response Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Evaluación holística y multi-metodológica de propuestas de seguridad en IA

Open in a new tab (opens in new tab)
Code (opens in new tab)
Share

holistic-eval

Evaluación holística y multi-metodológica de propuestas de seguridad en IA

Ante incidentes como el de OpenAI–Hugging Face, el cuello de botella no es la falta de propuestas, sino el tiempo disponible para evaluarlas con rigor desde varios ángulos.

holistic-eval es una herramienta de línea de comandos que evalúa un documento de propuesta frente a una clase de incidente mediante cuatro metodologías independientes ejecutadas en paralelo: prudencia, yellow teaming, red teaming y diseño sinérgico.

Un agregador, orquestado con LangGraph, reconcilia los cuatro informes en un veredicto único. La herramienta corre con Claude y aplica un tope de gasto duro y persistente.

En dos ejecuciones reales, produjo cuatro críticas diferenciadas y un veredicto reconciliado por aproximadamente US$0,40 por evaluación. El hallazgo principal es de viabilidad: resulta barato y reproducible obtener perspectivas metodológicamente distintas de forma automatizada.

Pruébenlo con propuestas del hackathon: github.com/mateo3264/holistic-eval

Uso:

uv sync cp .env.example .env

Agrega tu clave en el archivo .env local:

ANTHROPIC_API_KEY=sk-ant-

Luego ejecuta:

uv run holistic-eval su_propuesta.pdf

La herramienta acepta documentos en PDF, Markdown o texto.

No subas ANTHROPIC_API_KEY al repositorio, al README ni a un chat público. Solicita la clave al responsable del equipo por un canal privado y asegúrate de que .env esté incluido en .gitignore.

Por defecto usa Claude Haiku 4.5, una opción económica de aproximadamente US$0,40 por evaluación.

Para priorizar calidad, puedes cambiar a Sonnet 5 en .env:

LLM_MODEL=claude-sonnet-5 PRICE_PER_MTOK_INPUT=<precio publicado de entrada de Sonnet 5> PRICE_PER_MTOK_OUTPUT=<precio publicado de salida de Sonnet 5>

Es importante configurar PRICE_PER_MTOK_INPUT y PRICE_PER_MTOK_OUTPUT con las tarifas reales de Sonnet 5. El cálculo y cumplimiento del tope de gasto depende de esos valores; conservar las tarifas de Haiku (1/5) subestimaría el gasto.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical AI safety problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. Fairly interesting idea, but without any kind of output quality analysis, it's hard to judge how useful the eval is.

    Please next time for an international hackathon, submit the project in English as not everyone speaks Spanish.

Cite this project

@misc{acosta2026evaluacion,
  title = {{Evaluación holística y multi-metodológica de propuestas de seguridad en IA}},
  author = {Mateo Acosta},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/evaluacin-holstica-y-multimetodolgica-de-propuestas-de-seguridad-en-ia-j9pf}},
  url = {https://apartresearch.com/sprints/projects/evaluacin-holstica-y-multimetodolgica-de-propuestas-de-seguridad-en-ia-j9pf}
}

More from AI Incident Response Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026