Dataset Viewer
Auto-converted to Parquet Duplicate
suite
stringclasses
2 values
task_id
stringclasses
50 values
regime
stringclasses
2 values
category
stringclasses
10 values
policy
stringclasses
5 values
stage1_choice
stringclasses
4 values
stage1_correct
bool
2 classes
stage1_latency_s
float64
0.32
43.8
stage1_boost
bool
2 classes
stage2_choice
stringclasses
4 values
stage2_correct
bool
2 classes
stage2_latency_s
float64
0.26
25.5
workflow_total_latency_s
float64
0.78
53.8
workflow_both_stages_correct
bool
2 classes
quorumbench_50_3b_7b_14b
R01_sha256_bytes
routine
systems_basics
Fast_k_of_N
B
true
0.5153
false
B
true
0.4386
0.9539
true
quorumbench_50_3b_7b_14b
R01_sha256_bytes
routine
systems_basics
Sync_N_of_N
B
true
43.7576
false
B
true
10.0622
53.8198
true
quorumbench_50_3b_7b_14b
R01_sha256_bytes
routine
systems_basics
Genealogy_Sync_HPR
B
true
43.7576
false
B
true
10.0622
53.8198
true
quorumbench_50_3b_7b_14b
R01_sha256_bytes
routine
systems_basics
CVaR_Static_Cutoff
B
true
2.8
false
B
true
2.8
5.6
true
quorumbench_50_3b_7b_14b
R01_sha256_bytes
routine
systems_basics
QuorumGuard
B
true
2.6525
true
B
true
3.2562
5.9087
true
quorumbench_50_3b_7b_14b
T01_uint16_overflow
boundary_trap
systems_bugs
Fast_k_of_N
A
false
0.4363
false
B
true
0.5679
1.0042
false
quorumbench_50_3b_7b_14b
T01_uint16_overflow
boundary_trap
systems_bugs
Sync_N_of_N
A
false
14.3916
false
B
true
21.3475
35.7391
false
quorumbench_50_3b_7b_14b
T01_uint16_overflow
boundary_trap
systems_bugs
Genealogy_Sync_HPR
A
false
14.3916
false
B
true
21.3475
35.7391
false
quorumbench_50_3b_7b_14b
T01_uint16_overflow
boundary_trap
systems_bugs
CVaR_Static_Cutoff
A
false
2.8
false
A
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T01_uint16_overflow
boundary_trap
systems_bugs
QuorumGuard
A
false
4.9344
true
B
true
4.7779
9.7123
false
quorumbench_50_3b_7b_14b
T02_float32_precision_stall
boundary_trap
numerical
Fast_k_of_N
A
false
0.7703
false
A
false
0.6919
1.4622
false
quorumbench_50_3b_7b_14b
T02_float32_precision_stall
boundary_trap
numerical
Sync_N_of_N
A
false
24.8773
false
B
true
25.2597
50.137
false
quorumbench_50_3b_7b_14b
T02_float32_precision_stall
boundary_trap
numerical
Genealogy_Sync_HPR
B
true
24.8773
false
B
true
18.0793
42.9566
true
quorumbench_50_3b_7b_14b
T02_float32_precision_stall
boundary_trap
numerical
CVaR_Static_Cutoff
A
false
2.8
false
A
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T02_float32_precision_stall
boundary_trap
numerical
QuorumGuard
B
true
8.3037
true
B
true
6.143
14.4467
true
quorumbench_50_3b_7b_14b
R02_binary_heap_left_child
routine
algorithms
Fast_k_of_N
C
true
0.5036
false
B
false
0.4622
0.9658
false
quorumbench_50_3b_7b_14b
R02_binary_heap_left_child
routine
algorithms
Sync_N_of_N
C
true
18.2611
false
B
false
13.561
31.8221
false
quorumbench_50_3b_7b_14b
R02_binary_heap_left_child
routine
algorithms
Genealogy_Sync_HPR
C
true
18.2611
false
B
false
13.561
31.8221
false
quorumbench_50_3b_7b_14b
R02_binary_heap_left_child
routine
algorithms
CVaR_Static_Cutoff
C
true
2.8
false
B
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
R02_binary_heap_left_child
routine
algorithms
QuorumGuard
C
true
2.9477
true
B
false
5.0251
7.9728
false
quorumbench_50_3b_7b_14b
T03_sql_not_in_null
boundary_trap
databases
Fast_k_of_N
A
false
0.4458
false
B
true
0.3463
0.7921
false
quorumbench_50_3b_7b_14b
T03_sql_not_in_null
boundary_trap
databases
Sync_N_of_N
A
false
18.8742
false
B
true
14.8872
33.7614
false
quorumbench_50_3b_7b_14b
T03_sql_not_in_null
boundary_trap
databases
Genealogy_Sync_HPR
A
false
18.8742
false
B
true
14.8872
33.7614
false
quorumbench_50_3b_7b_14b
T03_sql_not_in_null
boundary_trap
databases
CVaR_Static_Cutoff
A
false
2.8
false
B
true
2.8
5.6
false
quorumbench_50_3b_7b_14b
T03_sql_not_in_null
boundary_trap
databases
QuorumGuard
A
false
3.0468
true
B
true
4.6125
7.6593
false
quorumbench_50_3b_7b_14b
T04_shift_add_precedence
boundary_trap
systems_bugs
Fast_k_of_N
A
false
0.7867
false
A
false
0.4379
1.2246
false
quorumbench_50_3b_7b_14b
T04_shift_add_precedence
boundary_trap
systems_bugs
Sync_N_of_N
B
true
16.6621
false
B
true
14.1764
30.8385
true
quorumbench_50_3b_7b_14b
T04_shift_add_precedence
boundary_trap
systems_bugs
Genealogy_Sync_HPR
B
true
16.6621
false
B
true
14.1764
30.8385
true
quorumbench_50_3b_7b_14b
T04_shift_add_precedence
boundary_trap
systems_bugs
CVaR_Static_Cutoff
A
false
2.8
false
A
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T04_shift_add_precedence
boundary_trap
systems_bugs
QuorumGuard
B
true
3.4066
true
B
true
4.9266
8.3332
true
quorumbench_50_3b_7b_14b
R03_http_idempotent
routine
distributed_systems
Fast_k_of_N
D
true
0.4163
false
A
false
0.5386
0.9549
false
quorumbench_50_3b_7b_14b
R03_http_idempotent
routine
distributed_systems
Sync_N_of_N
D
true
14.4318
false
B
true
13.4878
27.9196
true
quorumbench_50_3b_7b_14b
R03_http_idempotent
routine
distributed_systems
Genealogy_Sync_HPR
D
true
14.4318
false
B
true
13.4878
27.9196
true
quorumbench_50_3b_7b_14b
R03_http_idempotent
routine
distributed_systems
CVaR_Static_Cutoff
D
true
2.8
false
A
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
R03_http_idempotent
routine
distributed_systems
QuorumGuard
D
true
2.1804
true
B
true
4.8423
7.0227
true
quorumbench_50_3b_7b_14b
T05_python_xor_caret
boundary_trap
systems_bugs
Fast_k_of_N
A
false
0.5784
false
B
true
0.5331
1.1115
false
quorumbench_50_3b_7b_14b
T05_python_xor_caret
boundary_trap
systems_bugs
Sync_N_of_N
A
false
18.3921
false
B
true
13.8971
32.2892
false
quorumbench_50_3b_7b_14b
T05_python_xor_caret
boundary_trap
systems_bugs
Genealogy_Sync_HPR
B
true
18.3921
false
B
true
13.3539
31.746
true
quorumbench_50_3b_7b_14b
T05_python_xor_caret
boundary_trap
systems_bugs
CVaR_Static_Cutoff
A
false
2.8
false
B
true
2.8
5.6
false
quorumbench_50_3b_7b_14b
T05_python_xor_caret
boundary_trap
systems_bugs
QuorumGuard
B
true
6.1017
true
B
true
4.3626
10.4643
true
quorumbench_50_3b_7b_14b
T06_python_bankers_rounding
boundary_trap
numerical
Fast_k_of_N
A
false
0.5871
false
D
false
0.351
0.9381
false
quorumbench_50_3b_7b_14b
T06_python_bankers_rounding
boundary_trap
numerical
Sync_N_of_N
A
false
12.0262
false
B
true
18.9222
30.9484
false
quorumbench_50_3b_7b_14b
T06_python_bankers_rounding
boundary_trap
numerical
Genealogy_Sync_HPR
B
true
12.0262
false
B
true
19.4998
31.526
true
quorumbench_50_3b_7b_14b
T06_python_bankers_rounding
boundary_trap
numerical
CVaR_Static_Cutoff
A
false
2.8
false
D
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T06_python_bankers_rounding
boundary_trap
numerical
QuorumGuard
B
true
4.1618
true
B
true
6.4841
10.6459
true
quorumbench_50_3b_7b_14b
R04_matmul_flops
routine
numerical
Fast_k_of_N
A
true
0.7584
false
C
false
0.2643
1.0227
false
quorumbench_50_3b_7b_14b
R04_matmul_flops
routine
numerical
Sync_N_of_N
A
true
22.7883
false
C
false
14.3871
37.1754
false
quorumbench_50_3b_7b_14b
R04_matmul_flops
routine
numerical
Genealogy_Sync_HPR
A
true
22.7883
false
A
true
14.3871
37.1754
true
quorumbench_50_3b_7b_14b
R04_matmul_flops
routine
numerical
CVaR_Static_Cutoff
A
true
2.8
false
C
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
R04_matmul_flops
routine
numerical
QuorumGuard
A
true
2.8056
true
A
true
4.8333
7.6389
true
quorumbench_50_3b_7b_14b
T07_closure_late_binding
boundary_trap
systems_bugs
Fast_k_of_N
A
false
0.58
false
B
true
0.6237
1.2037
false
quorumbench_50_3b_7b_14b
T07_closure_late_binding
boundary_trap
systems_bugs
Sync_N_of_N
B
true
18.4878
false
B
true
11.9893
30.4771
true
quorumbench_50_3b_7b_14b
T07_closure_late_binding
boundary_trap
systems_bugs
Genealogy_Sync_HPR
B
true
18.4878
false
B
true
11.9893
30.4771
true
quorumbench_50_3b_7b_14b
T07_closure_late_binding
boundary_trap
systems_bugs
CVaR_Static_Cutoff
A
false
2.8
false
B
true
2.8
5.6
false
quorumbench_50_3b_7b_14b
T07_closure_late_binding
boundary_trap
systems_bugs
QuorumGuard
B
true
6.2441
true
B
true
4.4115
10.6556
true
quorumbench_50_3b_7b_14b
T08_shallow_list_multiply
boundary_trap
systems_bugs
Fast_k_of_N
A
false
0.796
false
B
true
0.3115
1.1075
false
quorumbench_50_3b_7b_14b
T08_shallow_list_multiply
boundary_trap
systems_bugs
Sync_N_of_N
A
false
15.5027
false
B
true
16.6079
32.1106
false
quorumbench_50_3b_7b_14b
T08_shallow_list_multiply
boundary_trap
systems_bugs
Genealogy_Sync_HPR
B
true
15.5027
false
B
true
14.7606
30.2633
true
quorumbench_50_3b_7b_14b
T08_shallow_list_multiply
boundary_trap
systems_bugs
CVaR_Static_Cutoff
A
false
2.8
false
B
true
2.8
5.6
false
quorumbench_50_3b_7b_14b
T08_shallow_list_multiply
boundary_trap
systems_bugs
QuorumGuard
B
true
5.556
true
B
true
4.2337
9.7897
true
quorumbench_50_3b_7b_14b
R05_tcp_syn_seq
routine
networking
Fast_k_of_N
C
true
0.5227
false
C
true
0.4732
0.9959
true
quorumbench_50_3b_7b_14b
R05_tcp_syn_seq
routine
networking
Sync_N_of_N
C
true
29.7119
false
C
true
18.3475
48.0594
true
quorumbench_50_3b_7b_14b
R05_tcp_syn_seq
routine
networking
Genealogy_Sync_HPR
C
true
29.7119
false
C
true
18.3475
48.0594
true
quorumbench_50_3b_7b_14b
R05_tcp_syn_seq
routine
networking
CVaR_Static_Cutoff
C
true
2.8
false
C
true
2.8
5.6
true
quorumbench_50_3b_7b_14b
R05_tcp_syn_seq
routine
networking
QuorumGuard
C
true
2.218
true
C
true
6.0589
8.2769
true
quorumbench_50_3b_7b_14b
T09_utf8_emoji_bytes
boundary_trap
systems_bugs
Fast_k_of_N
B
false
0.707
false
C
false
0.5264
1.2334
false
quorumbench_50_3b_7b_14b
T09_utf8_emoji_bytes
boundary_trap
systems_bugs
Sync_N_of_N
B
false
18.9767
false
C
false
13.527
32.5037
false
quorumbench_50_3b_7b_14b
T09_utf8_emoji_bytes
boundary_trap
systems_bugs
Genealogy_Sync_HPR
C
true
18.9767
false
B
true
13.4007
32.3774
true
quorumbench_50_3b_7b_14b
T09_utf8_emoji_bytes
boundary_trap
systems_bugs
CVaR_Static_Cutoff
B
false
2.8
false
C
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T09_utf8_emoji_bytes
boundary_trap
systems_bugs
QuorumGuard
C
true
4.7098
true
B
true
4.5037
9.2135
true
quorumbench_50_3b_7b_14b
T10_harmonic_throughput
boundary_trap
distributed_systems
Fast_k_of_N
D
false
0.7208
false
C
false
0.5804
1.3012
false
quorumbench_50_3b_7b_14b
T10_harmonic_throughput
boundary_trap
distributed_systems
Sync_N_of_N
B
true
18.0586
false
C
false
12.7001
30.7587
false
quorumbench_50_3b_7b_14b
T10_harmonic_throughput
boundary_trap
distributed_systems
Genealogy_Sync_HPR
B
true
18.0586
false
B
true
12.7001
30.7587
true
quorumbench_50_3b_7b_14b
T10_harmonic_throughput
boundary_trap
distributed_systems
CVaR_Static_Cutoff
D
false
2.8
false
C
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T10_harmonic_throughput
boundary_trap
distributed_systems
QuorumGuard
B
true
6.8345
true
B
true
4.4504
11.2849
true
quorumbench_50_3b_7b_14b
R06_raft_quorum_size
routine
distributed_systems
Fast_k_of_N
B
true
0.5884
false
C
false
0.4211
1.0095
false
quorumbench_50_3b_7b_14b
R06_raft_quorum_size
routine
distributed_systems
Sync_N_of_N
B
true
17.8698
false
C
false
18.0815
35.9513
false
quorumbench_50_3b_7b_14b
R06_raft_quorum_size
routine
distributed_systems
Genealogy_Sync_HPR
B
true
17.8698
false
A
false
18.0815
35.9513
false
quorumbench_50_3b_7b_14b
R06_raft_quorum_size
routine
distributed_systems
CVaR_Static_Cutoff
B
true
2.8
false
C
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
R06_raft_quorum_size
routine
distributed_systems
QuorumGuard
B
true
3.3373
true
A
false
6.0304
9.3677
false
quorumbench_50_3b_7b_14b
T11_bayes_false_positive
boundary_trap
math
Fast_k_of_N
A
false
0.3437
false
A
false
0.6346
0.9783
false
quorumbench_50_3b_7b_14b
T11_bayes_false_positive
boundary_trap
math
Sync_N_of_N
A
false
18.9539
false
A
false
14.9253
33.8792
false
quorumbench_50_3b_7b_14b
T11_bayes_false_positive
boundary_trap
math
Genealogy_Sync_HPR
B
false
18.9539
false
B
true
13.7631
32.717
false
quorumbench_50_3b_7b_14b
T11_bayes_false_positive
boundary_trap
math
CVaR_Static_Cutoff
A
false
2.8
false
A
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T11_bayes_false_positive
boundary_trap
math
QuorumGuard
B
false
7.0321
true
B
true
4.7579
11.79
false
quorumbench_50_3b_7b_14b
T12_int32_midpoint_overflow
boundary_trap
algorithms
Fast_k_of_N
B
true
0.6866
false
B
true
0.5108
1.1974
true
quorumbench_50_3b_7b_14b
T12_int32_midpoint_overflow
boundary_trap
algorithms
Sync_N_of_N
B
true
22.8047
false
B
true
14.3366
37.1413
true
quorumbench_50_3b_7b_14b
T12_int32_midpoint_overflow
boundary_trap
algorithms
Genealogy_Sync_HPR
B
true
22.8047
false
B
true
14.3366
37.1413
true
quorumbench_50_3b_7b_14b
T12_int32_midpoint_overflow
boundary_trap
algorithms
CVaR_Static_Cutoff
B
true
2.8
false
B
true
2.8
5.6
true
quorumbench_50_3b_7b_14b
T12_int32_midpoint_overflow
boundary_trap
algorithms
QuorumGuard
B
true
2.7252
true
B
true
5.1132
7.8384
true
quorumbench_50_3b_7b_14b
R07_dag_topo_cycle
routine
algorithms
Fast_k_of_N
B
true
0.5752
false
B
true
0.5182
1.0934
true
quorumbench_50_3b_7b_14b
R07_dag_topo_cycle
routine
algorithms
Sync_N_of_N
B
true
18.1268
false
B
true
9.7759
27.9027
true
quorumbench_50_3b_7b_14b
R07_dag_topo_cycle
routine
algorithms
Genealogy_Sync_HPR
B
true
18.1268
false
B
true
9.7759
27.9027
true
quorumbench_50_3b_7b_14b
R07_dag_topo_cycle
routine
algorithms
CVaR_Static_Cutoff
B
true
2.8
false
B
true
2.8
5.6
true
quorumbench_50_3b_7b_14b
R07_dag_topo_cycle
routine
algorithms
QuorumGuard
B
true
3.1991
true
B
true
3.4422
6.6413
true
quorumbench_50_3b_7b_14b
T13_sliding_window_fencepost
boundary_trap
algorithms
Fast_k_of_N
A
false
0.6701
false
D
false
0.6187
1.2888
false
quorumbench_50_3b_7b_14b
T13_sliding_window_fencepost
boundary_trap
algorithms
Sync_N_of_N
B
true
21.8993
false
C
false
13.2908
35.1901
false
quorumbench_50_3b_7b_14b
T13_sliding_window_fencepost
boundary_trap
algorithms
Genealogy_Sync_HPR
B
true
21.8993
false
C
false
13.2908
35.1901
false
quorumbench_50_3b_7b_14b
T13_sliding_window_fencepost
boundary_trap
algorithms
CVaR_Static_Cutoff
A
false
2.8
false
D
false
2.8
5.6
false
quorumbench_50_3b_7b_14b
T13_sliding_window_fencepost
boundary_trap
algorithms
QuorumGuard
B
true
5.9837
true
C
false
4.9969
10.9806
false
End of preview. Expand in Data Studio

QuorumBench v1.0

Benchmark tasks and Apple M5 Pro (48 GB unified memory, MPS FP16) telemetry for QuorumGuard: asynchronous multi-agent consensus across Qwen/Qwen2.5-3B-Instruct, Qwen/Qwen2.5-7B-Instruct, Qwen/Qwen2.5-Coder-14B-Instruct, HuggingFaceTB/SmolLM2-1.7B-Instruct, and ibm-granite/granite-3.1-8b-instruct with isolated Python sandbox verification on both QuorumBench and CRUXEval-50.

Dataset Configurations

  1. policy_decisions (400 rows): Stage-1, Stage-2, and end-to-end latency/accuracy outcomes across 5 core consensus policies (Fast_k_of_N, CVaR_Static_Cutoff, Sync_N_of_N, Genealogy_Sync_HPR, QuorumGuard) on both the 50-task 3-tier suite and 30-task 2-tier pilot (raw/revision_ablations_and_sweeps.json adds Fast_3_of_N_Maj, TW-Fast, Entropy-Cascade, 7-point CVaR sweep, 10-seed variances, 2,000-bootstrap 95% CIs, hardware-only latencies, and 4-axis/psi_i sensitivity).
  2. tasks (50 rows): Two-stage systems, numerical precision, SQL, and algorithmic verification workflows (20 routine tasks and 30 boundary traps) with shared upstream triage notes (raw/cruxeval50_crossfamily.json adds external CRUXEval-50 & SmolLM2-1.7B + Granite-3.1-8B runs).
  3. agent_observations (560 rows): Per-agent MPS wall-clock times, completion timestamps, pre-commitment 4-option Shannon entropies, top-2 logit margins, provenance roots, and Python sandbox execution receipts.
  4. gpu_contention (12 rows): Trace-driven 8-token KV-cached decode quantum contention on Qwen2.5-7B-Instruct across loads rho in {0.58, 0.72, 0.85, 0.93} (raw/live_mps_quantum_server.json adds the live 64-request concurrent multi-DynamicCache MPS server benchmark).

Summary Results (50 Two-Stage Workflows, 3B + 7B + 14B-Coder)

Policy S1 / S2 Acc Joint [95% CI] 10-Seed (Trap) Mean / P95 [95% CI] HW-Only P95
Fast_k_of_N (k=3) 46.0% / 52.0% 28.0% [16, 42] 27.8+-0.6% (6.7%) 1.18s / 1.49s [1.4, 1.6] 0.32s
Fast_3_of_N_Maj 46.0% / 52.0% 28.0% [16, 42] 28.0+-0.9% (6.7%) 1.34s / 1.58s [1.5, 1.6] 0.34s
CVaR_Cutoff (B=2.8s) 46.0% / 52.0% 28.0% [16, 42] 26.2+-1.1% (6.7%) 5.60s / 5.60s [5.6, 5.6] 5.60s
TW-Fast (Static Tier) 62.0% / 68.0% 42.0% [28, 56] 44.2+-3.0% (23.3%) 9.70s / 27.07s [19.7, 35.0] 16.05s
Entropy-Cascade 68.0% / 84.0% 58.0% [44, 72] 57.4+-2.3% (43.3%) 25.12s / 45.06s [40.5, 47.9] 28.81s
Sync_N_of_N 66.0% / 68.0% 42.0% [28, 56] 42.6+-1.9% (23.3%) 35.64s / 46.83s [43.4, 53.6] 30.19s
Genealogy_Sync_HPR 84.0% / 86.0% 72.0% [58, 84] 72.0+-1.6% (63.3%) 36.34s / 47.39s [43.7, 53.6] 30.19s
QuorumGuard 84.0% / 86.0% 72.0% [58, 84] 72.0+-1.6% (63.3%) 10.18s / 15.09s [13.8, 16.1] 13.24s

Pareto Frontier

Citation

@misc{syed2026quorumbench,
  author       = {Syed, Mohammed Kamran and Rentala, Anudeep and Rahul, Somesh and Pai, Raghuram Malpe and Dara, Krishna Suman},
  title        = {{QuorumBench}: Asynchronous Multi-Agent {LLM} Consensus and Straggler Herding Benchmark},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/batistado/QuorumBench}}
}
Downloads last month
60

Models trained or fine-tuned on batistado/QuorumBench

Space using batistado/QuorumBench 1