Research integrity
Evidence register & references
Every substantive claim in this platform is classified by what kind of statement it is. Nothing is upgraded from vendor material, and anything that cannot yet be supported is held as evidence required.
- Reference library
- 33 sources
- Classification
- Five categories
01 · Classification system
- Verified evidence
Published, openable research, or a direct measurement.
- Vendor material
Material supplied by the vendor. Not independently verified.
- Illustrative scenario
A business calculation built to illustrate, not a quotation from a source.
- Working hypothesis
An internal assumption made to be tested.
- Evidence required
A claim with no basis yet, which must be documented before anything rests on it.
02 · Verified scientific evidence
32 sources- March 2018World Models
David Ha, Jürgen Schmidhuber
arXiv:1803.10122 · NeurIPS 2018 workshop track
Open source ↗ - April 2025Mastering diverse control tasks through world models
Danijar Hafner, Jurgis Pasukonis, Jimmy Ba, Timothy Lillicrap
Nature 640, 647–653 (DreamerV3)
Open source ↗ - April 2023Generative Agents: Interactive Simulacra of Human Behavior
Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, Michael S. Bernstein
arXiv:2304.03442 · UIST 2023
Open source ↗ - June 2017Attention Is All You Need
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, Illia Polosukhin
arXiv:1706.03762 · NeurIPS 2017
Open source ↗ - May 2020Language Models are Few-Shot Learners
Tom B. Brown et al. (OpenAI)
arXiv:2005.14165 · NeurIPS 2020
Open source ↗ - January 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, Denny Zhou
arXiv:2201.11903 · NeurIPS 2022
Open source ↗ - March 2022Training language models to follow instructions with human feedback
Long Ouyang et al. (OpenAI)
arXiv:2203.02155 · NeurIPS 2022
Open source ↗ - October 2022ReAct: Synergizing Reasoning and Acting in Language Models
Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, Yuan Cao
arXiv:2210.03629 · ICLR 2023
Open source ↗ - May 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Patrick Lewis et al. (Facebook AI Research, University College London, New York University)
arXiv:2005.11401 · NeurIPS 2020
Open source ↗ - March 2023Reflexion: Language Agents with Verbal Reinforcement Learning
Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, Shunyu Yao
arXiv:2303.11366 · NeurIPS 2023
Open source ↗ - October 2023Large Language Models Cannot Self-Correct Reasoning Yet
Jie Huang, Xinyun Chen, Swaroop Mishra, Huaixiu Steven Zheng, Adams Wei Yu, Xinying Song, Denny Zhou
arXiv:2310.01798 · ICLR 2024
Open source ↗ - July 2022Language Models (Mostly) Know What They Know
Saurav Kadavath et al. (Anthropic)
arXiv:2207.05221
Open source ↗ - March 2022Self-Consistency Improves Chain of Thought Reasoning in Language Models
Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, Denny Zhou
arXiv:2203.11171 · ICLR 2023
Open source ↗ - May 2023Improving Factuality and Reasoning in Language Models through Multiagent Debate
Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, Igor Mordatch
arXiv:2305.14325 · ICML 2024
Open source ↗ - March 2017Overcoming catastrophic forgetting in neural networks
James Kirkpatrick et al. (DeepMind)
Proceedings of the National Academy of Sciences 114 (13)
Open source ↗ - June 2022
OpenReview · position paper
Open source ↗ - December 2018A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
David Silver, Thomas Hubert, Julian Schrittwieser et al. (DeepMind)
Science 362, 1140–1144
Open source ↗ - February 2023Toolformer: Language Models Can Teach Themselves to Use Tools
Timo Schick, Jane Dwivedi-Yu, Roberto Dessì et al. (Meta AI Research)
arXiv:2302.04761 · NeurIPS 2023
Open source ↗ - February 2023Augmented Language Models: a Survey
Grégoire Mialon, Roberto Dessì, Maria Lomeli et al. (Meta AI Research)
arXiv:2302.07842 · TMLR
Open source ↗ - January 2025
arXiv:2501.12948
Open source ↗ - December 2022Constitutional AI: Harmlessness from AI Feedback
Yuntao Bai et al. (Anthropic)
arXiv:2212.08073
Open source ↗ - 2018Reinforcement Learning: An Introduction (2nd edition)
Richard S. Sutton, Andrew G. Barto
MIT Press
Open source ↗ - October 2026Towards Financial World Modeling
Preprint (authors listed in the record)
arXiv:2610.09048
Open source ↗ - September 2024 · ICLR 2025MarS: A Financial Market Simulation Engine Powered by Generative Foundation Model
Junjie Li, Yang Liu, Weiqing Liu, Jiang Bian et al. (Microsoft Research Asia)
ICLR 2025 · arXiv:2409.07486
Open source ↗ - 2026MarketSim: Simulating Stock Markets with Large-Scale Generative Agents
Preprint (authors listed in the record)
ICML 2026 · arXiv:2410.22570
Open source ↗ - 2026Order Flow and Cryptocurrency Returns
Journal of Financial Markets
Journal of Financial Markets (Elsevier)
Open source ↗ - August 2026Do Deep Learning Methods Improve Financial Forecasts?
Office of Financial Research, U.S. Department of the Treasury
OFR working paper series
Open source ↗ - August 2025
arXiv:2508.02739
Open source ↗ - 2019Universal features of price formation in financial markets: perspectives from deep learning
Justin Sirignano, Rama Cont
Quantitative Finance 19 (9)
Open source ↗ - 2023Deep order flow imbalance: extracting alpha at multiple horizons from the limit order book
Petter N. Kolm, Jeremy Turiel, Nicholas Westray
Mathematical Finance 33 (4)
Open source ↗ - 2020Empirical Asset Pricing via Machine Learning
Shihao Gu, Bryan Kelly, Dacheng Xiu
The Review of Financial Studies 33 (5)
Open source ↗ - April 2019ABIDES: Towards High-Fidelity Market Simulation for AI Research
David Byrd, Maria Hybinette, Tucker Hybinette Balch
arXiv:1904.12066
Open source ↗
03 · Vendor-provided evidence
- December 2024
Vendor documentation
Open source ↗
What would change our view
An entry moves classification only when new evidence justifies it, and the reason is recorded. Assumptions that cannot be validated within the research window remain assumptions, and are treated as such in every subsequent decision.
- A vendor quotation with comparable scope, allowing a like-for-like total-cost comparison.
- A measured rollout-accuracy curve for a world model on data it has never seen.
- A proof of concept that fails its pre-agreed capability threshold, reported honestly.
- Evidence that the relevant capability is commoditising faster than a licence could deliver it.
Sources consulted: published research literature; public protocol specifications; provider documentation reviewed at an introductory level only. No formal vendor evaluation has taken place.
Open the strategic assessment →