Question Answering on HotpotQA Modified (test) benchmark leaderboard · SOTA2 Research

Question Answering on HotpotQA Modified (test)

42.04AccuracyCorrectness-only

Evaluation Results

Method	Links
Correctness-only2026.07		42.04	0.5292	-0.1282	0.5486	0.5508	-0.547
Overconfidence-12026.07		42.04	0.7332	0.2012	0.2192	0.1182	-0.1152
Log- 1 / ln 2022026.07		41.86	0.7081	0.1922	0.2264	0.1038	-0.1026
Underconfidence-12026.07		41.81	0.7519	0.214	0.2041	0.0794	-0.0775
Brier-12026.07		41.73	0.7474	0.2131	0.2043	0.0637	-0.0608
Overconfidence-42026.07		41.66	0.7405	0.2149	0.2016	0.0223	-0.0129
Overconfidence-10002026.07		41.35	0.6856	0.1402	0.2734	0.23	-0.229
Underconfidence-42026.07		41.18	0.7481	0.2088	0.203	0.0616	-0.0565
Brier-log Hybrid2026.07		40.07	0.744	0.1994	0.2013	0.0503	-0.0459
Brier-22026.07		28.62	0.89	0.1677	0.1185	0.0414	-0.0414
Base (JSON)2026.07		28.06	0.6559	-0.2226	0.5032	0.5335	-0.5317
SFT (XML)2026.07		27.87	0.6472	-0.227	0.5057	0.5347	-0.5325
Log-12026.07		9.5	0.9681	0.0536	0.0415	0.0235	-0.0235
Base (XML)2026.07		3.63	0.605	-0.6437	0.68	0.7703	-0.7703
Log Loss2026.07		0	—	0	0.0001	0.005	-0.005
Brier Score2026.07		0	—	0	0.0001	0.005	-0.005