grouprec — leaderboard
Generated 2026-07-13 09:47 UTC ·
deep models and aggregators compared on the same data, splits, and metrics.
A. Coupled (sampled group-level HR/NDCG) — deep models vs aggregators
Faithful reimplementations on the sampled 1-vs-N protocol, fixed-budget training
(no best-epoch test-peeking). Numbers land in published ranges (e.g. Mafengwo ConsRec ~0.91 HR@5,
AlignGroup ~0.86; CAMRa2011 ConsRec ~0.6); residual gaps vs. paper headlines come from their
best-epoch-over-many-configs selection.
| dataset |
recommender |
HR@5 |
NDCG@5 |
HR@10 |
NDCG@10 |
| CAMRa2011 |
EASE+AVG |
0.5545 |
0.3533 |
0.7745 |
0.4253 |
| CAMRa2011 |
EASE+GFAR |
0.5869 |
0.3886 |
0.7966 |
0.4569 |
| CAMRa2011 |
NCF |
0.6179 |
0.4138 |
0.8131 |
0.4777 |
| CAMRa2011 |
AGREE |
0.6310 |
0.4303 |
0.8110 |
0.4892 |
| CAMRa2011 |
GroupIM |
0.6297 |
0.4268 |
0.8124 |
0.4867 |
| CAMRa2011 |
HyperGroup |
0.5924 |
0.4027 |
0.7972 |
0.4692 |
| CAMRa2011 |
ConsRec |
0.6262 |
0.4026 |
0.8124 |
0.4635 |
| CAMRa2011 |
AlignGroup |
0.5890 |
0.3868 |
0.7979 |
0.4548 |
| Mafengwo |
EASE+AVG |
0.1950 |
0.1550 |
0.2563 |
0.1749 |
| Mafengwo |
EASE+GFAR |
0.4030 |
0.2896 |
0.6563 |
0.3699 |
| Mafengwo |
NCF |
0.6633 |
0.5082 |
0.7910 |
0.5502 |
| Mafengwo |
AGREE |
0.6683 |
0.5159 |
0.7889 |
0.5551 |
| Mafengwo |
GroupIM |
0.6372 |
0.5019 |
0.7286 |
0.5316 |
| Mafengwo |
HyperGroup |
0.6794 |
0.5052 |
0.7950 |
0.5432 |
| Mafengwo |
ConsRec |
0.9317 |
0.9317 |
0.9317 |
0.9317 |
| Mafengwo |
AlignGroup |
0.8784 |
0.6947 |
0.9779 |
0.7278 |
B. Decoupled (MovieLens) — aggregator relevance vs fairness trade-off
Higher ndcg.min/minmax = fairer; LTP/RLProp lead on
fairness while AVG leads raw utility.
| dataset |
recommender |
ndcg.mean (utility) |
ndcg.min (fairness) |
ndcg.minmax (balance) |
| ml-100k |
LTP |
0.3295 |
0.2669 |
0.6902 |
| ml-100k |
RLProp |
0.3295 |
0.2669 |
0.6902 |
| ml-100k |
EPFuzzDA |
0.3295 |
0.2669 |
0.6902 |
| ml-100k |
GreedyLM |
0.3221 |
0.2616 |
0.6846 |
| ml-100k |
AVG |
0.3488 |
0.2166 |
0.4605 |
| ml-100k |
LMS |
0.2698 |
0.1925 |
0.5518 |
| ml-100k |
GFAR |
0.2730 |
0.1227 |
0.2802 |