feat: add lightweight GraphRAG benchmark - #77

Closed

Postroggy wants to merge 18 commits into

hugegraph:mainfrom

Postroggy:feat/graphrag-benchmark-issue7

Commits on Jul 2, 2026

feat: graphrag benchmark
Postroggy
committed
801db09

Commits on Jul 5, 2026

feat(benchmark): reproducible LLM-Judge client and metric direction registry
Postroggy
committed
8d02321
feat(benchmark): add GraphExtractFlow normalization and fail-fast doc-ID validation
Postroggy
committed
782a597
local: keep experimental scripts, docs and pipeline changes (not for upstream)
Postroggy
committed
8a6bcb7
feat(benchmark): add extraction adapter and schema normalization
Postroggy
committed
b97d083
feat(benchmark): add retrieval context adapter from WkFlowState
Postroggy
committed
2689732
feat(benchmark): include failed and degraded samples in Markdown report
Postroggy
committed
6447d1f
refactor(benchmark): make degraded-sample detection mode-aware
Postroggy
committed
f6e9b64
refactor(benchmark): rename degraded samples to low-performing samples
Postroggy
committed
448d44d
fix(benchmark): remove misleading low-performing section and add metric direction indicators
Postroggy
committed
a0abdfb
revert(local): remove non-benchmark scripts/docs and restore other modules to origin/main
Postroggy
committed
209c388
style(benchmark): fix ruff import sorting in graph_extract and retrieval_adapter
Postroggy
committed
930a149
Revert "revert(local): remove non-benchmark scripts/docs and restore other modules to origin/main"
Postroggy
committed
60b695a
revert(cleanup): undo the accidental revert and restore benchmark-only scope
Postroggy
committed
bb750d0
feat(benchmark): analytical markdown reporter with dimension roll-up
Postroggy
committed
dca5b3d
feat(benchmark): add jitter_baseline.py for demo compare
Postroggy
committed
f616f48

Navigation Menu

Uh oh!

feat: add lightweight GraphRAG benchmark - #77

feat: add lightweight GraphRAG benchmark#77
Postroggy wants to merge 18 commits into
hugegraph:mainfrom
Postroggy:feat/graphrag-benchmark-issue7

Commits on Jul 2, 2026

feat: graphrag benchmark

Commits on Jul 5, 2026

feat(benchmark): reproducible LLM-Judge client and metric direction registry

feat(benchmark): add GraphExtractFlow normalization and fail-fast doc-ID validation

local: keep experimental scripts, docs and pipeline changes (not for upstream)

feat(benchmark): add extraction adapter and schema normalization

feat(benchmark): add retrieval context adapter from WkFlowState

feat(benchmark): include failed and degraded samples in Markdown report

refactor(benchmark): make degraded-sample detection mode-aware

refactor(benchmark): rename degraded samples to low-performing samples

fix(benchmark): remove misleading low-performing section and add metric direction indicators

revert(local): remove non-benchmark scripts/docs and restore other modules to origin/main

style(benchmark): fix ruff import sorting in graph_extract and retrieval_adapter

Revert "revert(local): remove non-benchmark scripts/docs and restore other modules to origin/main"

revert(cleanup): undo the accidental revert and restore benchmark-only scope

feat(benchmark): analytical markdown reporter with dimension roll-up

feat(benchmark): add jitter_baseline.py for demo compare

Commits on Jul 6, 2026

fix(benchmark): improve retrieval data pipeline and syntax_validity behavior

feat(benchmark): add LLM-based extraction metrics (SemanticEntityF1, SemanticTripleF1, ExtractionFaithfulness)