53.00 H²EPRScore
The leading system still leaves substantial headroom on complete event-process reconstruction.
H²EPR-Bench
An evidence-traceable benchmark for event-process reconstruction.
Given the same fixed, multi-source evidence, can a model reconstruct how a complex event actually unfolded? H²EPR-Bench evaluates the process structure behind the facts.
3,000event instances
6domains
26event categories
25,814verified documents
3,000reference EPGs
21LLMs evaluated
Key Findings
Facts and coarse order
Source attribution and stage-order scores exceed 82, showing that models often identify the right material.
Relations remain below 19
Action–outcome, causal, and explicit temporal relations are much harder to reconstruct.
9.83% path retention
Mediated mechanism paths are rarely preserved, pinpointing the gap between event facts and event logic.
Background & Motivation
Real-world events are not static text objects. They unfold through stages, actors, actions, dependencies, outcomes, and evidence distributed across sources. Evaluating whether a model understands such events therefore requires more than summaries, timelines, or isolated extraction targets.
Why process graphs?
Event-process graphs keep stage order, actor interactions, operations, relations, and evidence links visible in the same representation.
This turns event understanding into a structured reconstruction problem rather than a single text-generation target.
H²EPR-Bench treats events as structured process objects. Starting from fixed evidence contexts, each event is represented as a hierarchical heterogeneous Event-Process Graph that makes stage progression, actor interactions, temporal dependencies, action-result logic, and evidence links explicit.
H²EPR-Bench centers on reconstruction, while the EPG representation also creates a common substrate for cross-event analysis, continuation, prediction, and simulation research.
Event graphs
→
Linked processes
→
Scenario simulation
→
Societal world model
Benchmark Construction
Task
Fixed-evidence reconstruction
Each system receives an event descriptor and the same fixed evidence package, then returns a structured Event-Process Graph.
Reference
Expert-finalized references
Agent–expert review, human adjudication, and final sign-off produce one consistent medium-granularity reference EPG per event.
Graph
Hierarchical heterogeneous representation
The target representation connects macro stages, meso episodes, micro-level participants, actions, outcomes, relations, and evidence support.
Scoring
Diagnostic graph matching
Evaluation reports structural, temporal, causal, and evidence fidelity, together with a unified H²EPRScore.
FinMycelium Draft EPG
broad process coverage
→
Expert finalization
consistent evaluation reference
FinMycelium enables graph construction at scale; expert review then aligns scope, process relevance, and evidence support before each reference EPG is finalized.
Dataset Overview
H²EPR-Bench covers 3,000 events from 1629 to 2025 across six domains and twenty-six categories. Its evidence packages retain 25,814 verified documents from 84,693 retrieved source records, with every artifact aligned by a stable event ID.
Interactive Explorer
The Explorer is the browsing layer for event metadata, stage tables, timeline views, and public graph summaries.
Inspect events without navigating raw files.
Filter the catalog, open an event profile, inspect stage rows, view timelines, and jump to public artifacts.
Explore Events
Representative Event Processes
Representative FinMycelium Draft-EPG timelines from the 3,000-event collection.
Results & Diagnostics
Direct reconstruction results expose a consistent gap between evidence retention and structured event-process recovery.
01
Fixed evidence
Each system receives the same event descriptor and evidence context.
02
Graph output
The model returns a hierarchical heterogeneous event-process graph.
03
Schema gate
Every benchmark instance contributes to the official score after schema normalization.
04
Graph matching
Outputs are aligned with expert-finalized reference EPGs.
05
Diagnostics
Scores separate structural, temporal, causal, and evidence fidelity.
21
language models evaluated
3,000
events per system
53.00
best H²EPRScore
14 / 21
systems above 90% valid outputs
Schema validity is not enough.
Even mostly valid graph outputs remain weak on causal and mechanism reconstruction.
Model traces
Each card summarizes one direct reconstruction baseline across the official score and diagnostic subscores.
Loading model diagnostics...
Diagnostic views
Instance-level spread anchors the diagnostic panel; aggregate readouts and companion checks sit beside and below it.
Diagnostic readout
Aggregate scores hide the bottleneck.
The mean fidelity profile separates source support from process reconstruction. Causal and mechanism recovery remains the most pronounced bottleneck.
Full numeric table
Sortable all-instance results for 21 direct reconstruction systems across 3,000 events.
| System | Valid outputs (%) | H²EPRScore | Absolute | Structure | Temporal | Causal | Evidence |
|---|---|---|---|---|---|---|---|
| Loading leaderboard data... | |||||||
Access
Event Explorer
Interactive browsing across all 3,000 events and their public Draft EPGs.
Public Dataset
Metadata, public tables, sanitized Draft EPG artifacts, Gantt views, schemas, and manifests.
FinMycelium System
The multi-agent event reconstruction system used to construct public Draft EPG artifacts.
Reference EPGs (Gated)
Official medium-granularity reference EPGs for scoring and controlled benchmark use.
Release Repository
Public source for the website, Explorer, schemas, validators, and release tooling.
Citation
@misc{h2eprbench2026,
title = {H²EPR-Bench: An Evidence-Traceable Benchmark for Event-Process Reconstruction},
author = {AgenticFinLab},
year = {2026},
note = {Forthcoming}
}