State estimation under partial observation
Infer latent quantitative state from incomplete, noisy or delayed observations without inventing unobserved certainty.
REPRESENTATIONUNCERTAINTYPREDICTIONCross-domain evaluation / LQBENCH-QGI/0.7
Generality begins where task-family familiarity ends.
R4 turns the evaluation design into a public demonstration contract: task manifests, held-out transfer, evidence packets, failure records and explicit resource budgets.
Task families
Eight quantitative problem families designed to travel across domains.
Forecasting under regime shift
Predict beyond the calibration regime and identify when distribution shift invalidates prior assumptions.
PREDICTIONUNCERTAINTYTRANSFERConstrained optimisation
Optimise an explicit objective while respecting hard constraints, feasibility and reproducible solver conditions.
MODELINGOPTIMIZATIONEVIDENCEInverse problem reconstruction
Recover plausible hidden parameters or structures from observable consequences and report non-identifiability.
REPRESENTATIONMODELINGUNCERTAINTYSimulator-guided control
Choose bounded actions in a dynamic simulator while preserving state, constraints and stop conditions.
SIMULATIONOPTIMIZATIONAUTONOMYExperiment selection
Select the next measurement or simulation to reduce uncertainty or discriminate between competing hypotheses.
EXPERIMENT_DESIGNUNCERTAINTYMEMORYMultifidelity transfer
Compose cheap and expensive quantitative engines while preserving error bounds and domain validity.
SIMULATIONTRANSFERMODELINGEvidence and provenance audit
Reconstruct how a quantitative conclusion was produced and fail closed when lineage, runtime or uncertainty evidence is missing.
EVIDENCEMEMORYAUTONOMYCross-domain matrix
One task family. Different quantitative worlds.
A QGI claim should survive materially different system dynamics, data structures, constraints and sources of uncertainty.
Task familyStochastic & economic systemsPhysical & dynamical systemsEngineering & design systemsMolecular & biological systemsMaterials & scientific systems
TF-01●●●●●
TF-02●●●●●
TF-03●●●●●
TF-04●●●●●
TF-05●●●●●
TF-06●●●●●
TF-07●●●●●
TF-08●●●●●
A filled cell means the protocol permits a task instance in that domain. It is not a claim that LargeQuant or another system has passed that cell.
Evaluation execution
From pre-registered task to evidence-bearing result.
01Pre-register
→Freeze task manifest, tools, metrics, budgets and failure rules.
02Run baselines
→Execute reference systems under the same task contract.
03Run candidate
→Use matched seeds or instances where stochasticity matters.
04Verify evidence
→Check outputs, uncertainty, resources, provenance and reproducibility.
05Publish vector
Report domain-native metrics and capability dimensions, not a magic number.
Evaluation manifest
What was evaluated?
evaluation_idprotocol_versionsystem_idsystem_versiontask_manifest_hashbaseline_idsheld_out_regimerun_countresource_budgetenvironment_hashevidence_bundle_idstarted_atcompleted_atResult record
What actually happened?
evaluation_idtask_idtask_versiondomaindimension_vectorprimary_metricsecondary_metricsuncertainty_metricsbaseline_relative_metricsresource_metricsstatusfailure_reasonevidence_hashR4 creates the publication path for evidence-backed QGI demonstrations. It still does not fabricate results.
Public result arrays remain empty until benchmark tasks are executed under this protocol.
Demonstration protocol