Datasets:

TonyChen06
/

AscendingHarmonicReasoningInstruction

Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0)

35c6ff7 verified about 2 months ago

1.1
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
1.2
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
1.3
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
1.4
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
1.5
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
2.1
Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0) about 2 months ago
2.2
Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0) about 2 months ago
2.3
Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0) about 2 months ago
2.4
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
2.5
Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0) about 2 months ago
3.1
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
3.2
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
3.3
Redesign task 3.3 (temporal lag): wave-packet bursts so 'which is earlier' is well-posed (was ambiguous phase-shift) about 2 months ago
3.4
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
3.5
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
4.1
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
4.2
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
4.3
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
4.4
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
4.5
Fix task 4.5 (event clustering): randomize uniform-class positions so dedup keeps both classes balanced about 2 months ago
5.1
Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0) about 2 months ago
5.2
Regression/measurement tasks (2.1,2.2,2.3,2.5,5.1,5.2): hold out nuisances not the target, so answers stay in training support (exact-match no longer structurally 0) about 2 months ago
5.3
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
5.4
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
5.5
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago
5.6
Ahri v2.0: 26-task discrete-grid dataset (5000/1667/1667, deduplicated) about 2 months ago