Back to the episode map

Research Note

Multimodal Claim Evidence Ladder

| State | What it establishes | |---|---| | Concept | A proposed architecture, task, or research direction | | Internal experiment | A tested setup under reported conditi

Aug 4, 20261 min readBy Dalton Anderson

Multimodal Claim Evidence Ladder

Evidence states

StateWhat it establishes
ConceptA proposed architecture, task, or research direction
Internal experimentA tested setup under reported conditions
Benchmark resultA measured result for a named artifact, dataset, metric, and baseline
DemonstrationA selected interaction or prototype path
Released artifactWeights, code, or another reproducible object with terms
Accessible interfaceAn API or product surface available to a defined audience
Supported capabilityA documented, maintained feature with operational boundaries

Claim record

Record exact input and output modalities, task, model path, component versions, training and evaluation data, metric, comparator, limitation, artifact identity, access, terms, product surface, support state, geography, account, and verification date.

Historical application

The Llama 3 paper reported compositional image, video, and speech experiments. Its research page explicitly said the resulting models were still under development and not broadly released. The July 2024 Llama 3.1 model card identified the released family as text in and text out.

Later multimodal Llama releases are descendant evidence. They do not make the 2024 text artifacts multimodal retroactively.

Sources

Follow the evidence.

  1. ai.meta.com: the llama 3 herd of modelsai.meta.com
  2. ai-challenges.nist.gov: genaiai-challenges.nist.gov
  3. owasp.org: www project top 10 for large language model applicationsowasp.org
  4. youtu.be: 1KNOcY e9Tsyoutu.be
  5. github.com: PurpleLlamagithub.com
  6. crfm.stanford.edu: indexcrfm.stanford.edu
  7. NIST AI Risk Management Frameworknist.gov
  8. mlcommons.org: jailbreak 0 7mlcommons.org
  9. mlcommons.org: safety faqmlcommons.org
  10. github.com: MODEL CARDgithub.com
  11. ai-challenges.nist.gov: ariaai-challenges.nist.gov
  12. github.com: MODEL CARDgithub.com
  13. daltonanderson.ghost.io: metas llama 3 safety scaling and simple solutionsdaltonanderson.ghost.io
  14. ai.meta.com: meta llama 3 1 ai responsibilityai.meta.com
  15. NIST Generative AI Profilenvlpubs.nist.gov
  16. mlcommons.org: safety methodologymlcommons.org
  17. huggingface.co: concept guidehuggingface.co
  18. github.com: MODEL CARDgithub.com
  19. csrc.nist.gov: red teamingcsrc.nist.gov
  20. open.spotify.com: 44o5OPSumaZJcvRkXutorBopen.spotify.com
Multimodal Claim Evidence Ladder