agent-precog.org

GAP MATRIX · REGENERATED FROM 23 PAPERS

What the literature covers, and what PreCog needs

Rows are the five pillars. Columns are the properties PreCog's architecture requires. Each cell counts the included papers whose extraction note tags that property within that pillar; click a cell to see them. Empty cells are the research gaps.

Coverage strong (8+)partial (3–7)thin (1–2)gap (0)

Pillar	Multi-agent state	Non-generative inference	Latency / jitter aware	Time-to-irreversibility	Edge deployment	Human handoff	Cross-domain evaluation
1 · Multi-agent systems 2 papers	2	0	0	0	0	0	0
2 · World models 6 papers	2	6	0	0	0	0	0
3 · Learning to defer 9 papers	0	8	0	2	0	1	0
4 · Edge inference 0 papers	0	0	0	0	0	0	0
5 · Human–automation 6 papers	0	0	0	0	0	6	0

GAP

No multi-agent systems paper on non-generative inference

Pillar 1 has no included paper tagged with this requirement. Either the literature is silent or the search has not yet found it.

GAP

No multi-agent systems paper on latency / jitter aware

Pillar 1 has no included paper tagged with this requirement. Either the literature is silent or the search has not yet found it.

GAP

No multi-agent systems paper on time-to-irreversibility

Pillar 1 has no included paper tagged with this requirement. Either the literature is silent or the search has not yet found it.

All 28 empty cells

Pillar 1 × Non-generative inference
Pillar 1 × Latency / jitter aware
Pillar 1 × Time-to-irreversibility
Pillar 1 × Edge deployment
Pillar 1 × Human handoff
Pillar 1 × Cross-domain evaluation
Pillar 2 × Latency / jitter aware
Pillar 2 × Time-to-irreversibility
Pillar 2 × Edge deployment
Pillar 2 × Human handoff
Pillar 2 × Cross-domain evaluation
Pillar 3 × Multi-agent state
Pillar 3 × Latency / jitter aware
Pillar 3 × Edge deployment
Pillar 3 × Cross-domain evaluation
Pillar 4 × Multi-agent state
Pillar 4 × Non-generative inference
Pillar 4 × Latency / jitter aware
Pillar 4 × Time-to-irreversibility
Pillar 4 × Edge deployment
Pillar 4 × Human handoff
Pillar 4 × Cross-domain evaluation
Pillar 5 × Multi-agent state
Pillar 5 × Non-generative inference
Pillar 5 × Latency / jitter aware
Pillar 5 × Time-to-irreversibility
Pillar 5 × Edge deployment
Pillar 5 × Cross-domain evaluation

PILLAR 1 × MULTI-AGENT STATE · 2

A concise introduction to decentralized POMDPs (Springer 2016)
The complexity of decentralized control of Markov decision processes (Mathematics of Operations Research 2002)

PILLAR 2 × MULTI-AGENT STATE · 2

Graph-level representation learning with joint-embedding predictive architectures (TMLR 2025)
Hyperbolic graph convolutional neural networks (NeurIPS 2019)

PILLAR 2 × NON-GENERATIVE INFERENCE · 6

Graph-level representation learning with joint-embedding predictive architectures (TMLR 2025)
Self-supervised learning from images with a joint-embedding predictive architecture (CVPR 2023)
A path towards autonomous machine intelligence (OpenReview 2022)
Hyperbolic graph convolutional neural networks (NeurIPS 2019)
Poincaré embeddings for learning hierarchical representations (NeurIPS 2017)
VICReg: variance-invariance-covariance regularization for self-supervised learning (ICLR 2022)

PILLAR 3 × NON-GENERATIVE INFERENCE · 8

Two-stage learning to defer with multiple experts (NeurIPS 2023)
Learning-to-defer for sequential medical decision-making under uncertainty (arXiv 2021)
Consistent estimators for learning to defer to an expert (ICML 2020)
Calibrated learning to defer with one-vs-all classifiers (ICML 2022)
Predict responsibly: improving fairness and accuracy by learning to defer (NeurIPS 2018)
Safe reinforcement learning via shielding (AAAI 2018)
Selective classification for deep neural networks (NeurIPS 2017)
On optimum recognition error and reject tradeoff (IEEE Trans. Information Theory 1970)

PILLAR 3 × TIME-TO-IRREVERSIBILITY · 2

Learning-to-defer for sequential medical decision-making under uncertainty (arXiv 2021)
Control barrier function based quadratic programs for safety critical systems (IEEE Trans. Automatic Control 2017)

PILLAR 3 × HUMAN HANDOFF · 1

Predict responsibly: improving fairness and accuracy by learning to defer (NeurIPS 2018)

PILLAR 5 × HUMAN HANDOFF · 6

Ironies of artificial intelligence (Ergonomics 2023)
Moral crumple zones: cautionary tales in human-robot interaction (Engaging Science Technology and Society 2019)
Measurement of situation awareness in dynamic systems (Human Factors 1995)
Ironies of automation (Automatica 1983)
Trust in automation: designing for appropriate reliance (Human Factors 2004)
A model for types and levels of human interaction with automation (IEEE Trans. SMC-A 2000)