Registro de veracidade, rodada de 25/09/2026. Registro LUA, rodado pela LUA Vision.
Confira cada arquivo com: shasum -a 256 <arquivo>

Arquivos deste pacote:
94d431ed78b3694d92d30e2e71f41877b07be9fcc8edd64898967b69a411e324  benches/_comum.mjs
e0c2b8c40fcc49f0d157d10e604d034c660cb476d8cb5dd521c706a361224649  benches/_recusa.mjs
e826c1f0c79c11dace2409362532fe4c7476f7c66438fb117fe444e27f76272c  benches/orbench_hard.mjs
cd413e27b42bb76b85ab0608425a32b1e88b7986e1c463650ba833c9bcf3ede0  benches/orbench_toxic.mjs
2091130d9bd8e4b2e69ac9e1e5d0305fe11d7c2215ce2e3cf10805689314d295  conjuntos/hallulens_fonte.json
c2c11d849a74f550eaa14b3f06de0b6f855cc796fb566b4263c178544a34a5d5  conjuntos/hallulens_mixed_semente0_n200.jsonl
f999daa20586d09d981fc727d6f09e0bde72e59ad2735be87f33e18f73134ea7  preparar_abstention.py
29e6c351c0c279ca132d72e2f54be54f3a66a7f126610d11bde4e4d2f1628124  preparar_dados.py
09926f470ea2723ea5932da35a88df9f371f6d68a3a44f45e0638970499f4729  preparar_hallulens.py
93bd7df71c3a2fa3a5d9ea01ee767dca7583eb2cd7828c375458cd38c33312ac  rejulgar.mjs
07c0a61d9431fa441f3c1acf2415028f9aee7c46b607dce375aa5760d64e2faf  resultados-agregados.json
23080f09ca2e053a61a3a270d950a04729e8615bb5454a0f71b3b1061cc931a3  run.mjs
9a0252fa5a6ad6997373ddef3b3177f1406847a1623117dc2f4eb173ea89231e  tabela.mjs

Resultados por item (fora deste pacote; o SHA-256 é o do tabela.json da rodada):
d6ea6160050360e74d906e6163e71c9b4aa2b658383cd931b15d7b37d4f78ba5  resultados/abstention.house.jsonl
0ace3bda4aef2f948b5bebb878d5452913dd68055aa5cd7fece45f23869325ca  resultados/hallulens.house.jsonl
cb3a6155906961f51ef866c53b7beeec34b287b8414681ff36457d07a8419b0e  resultados/omniscience.house.jsonl
bb54212a3ab0f35aff6ddbd7b6ea1912216bad8d68a0f8887bbcc369a2a2a8cf  resultados/orbench_hard.house.jsonl
fc3d192d7e96e571d75dd28c1f42834b453de228e91159e568264fe4a9af1481  resultados/orbench_hard_j2.house.jsonl
dc847d11cf206365788318cf0cfd247dcdfe65cb855bca9eeca2f4ba2bc38972  resultados/orbench_toxic.house.jsonl
c188224d283e52280f4151307dc5c35b720177d454b51c082c62449df991c6c4  resultados/orbench_toxic_j2.house.jsonl
4bf96842d6d64026ea3b62f35d0158056af4dc4fe0fa3e36d411922f7cf48492  resultados/simpleqa.house.jsonl
dfe064c29b9fa9fec347a38365a3fa9e8c7a24606c84cd3f5557e3b1d1fd2969  resultados/simpleqa_conf.house.jsonl
f4682f8be975809e540ce439d060d5346674588754affe3fa9a7cb05c36e4ff9  resultados/xstest.house.jsonl

Resultados por item da comparação entre modelos (fora deste pacote):
0230d79a74260a4aced7d29571850a79970fe73bd2326485f6497231a099bbc2  resultados/abstention.house.rejulgado.jsonl
7c88a9f9e4875bc3cff1d4ecb15d49f481dfc2894746d54da47e70e08eef16d7  resultados/abstention.gpt55.jsonl
8de18028f01b92631ee3488f0728fb18f8d87ec5c27e714563b0353d0d0cc6ff  resultados/abstention.gpt54.jsonl
6bb01490b6656a6ff07a98d0a5d81df4f19c3ea744cc82c107fa6f622fa781d5  resultados/abstention.grok.jsonl
dceef4950539caa8a63705511bb0fd4547f27006409994fb18200ff058ffb0bb  resultados/abstention.deepseek.jsonl
38b2993bdd4919d40e310ccd8797669fad3a5f4d3e0e6bf595ba7aafca8b4afc  resultados/abstention.kimi.jsonl
6f7890d3939ad22423bd51f9e89ff54e02d4f641e3bf8da691c0ea01ff08b2ab  resultados/hallulens.house.rejulgado.jsonl
31174c8f7c0343ab16cf55ed2a42450dc0fe5210826ddad6cd345dd1e3b597c3  resultados/hallulens.gpt55.jsonl
3cd3c0ec7de9881a56f473957b5425198cfffda237f26957e84fe8e9ffa9d2d8  resultados/hallulens.gpt54.jsonl
3d92d38fbe2f7f08c056edc1833cc32d67cb8daf4c582e1b1357036cb05df8fd  resultados/hallulens.grok.jsonl
cd209cb0cc726b12dececfb8049992e18ee5a447f30f0b5f5d6001114f8fac06  resultados/hallulens.deepseek.jsonl
34b386c4e130b7888718eab9e1cd9189a94bb66c3b55aaf5b806219aa5b6f479  resultados/hallulens.kimi.jsonl
083505b520d6fe6dc77a57600f7855d1b7c30a42a836ab38792fbcb0e1d675c9  resultados/omniscience.house.rejulgado.jsonl
cd661e721d0e8d784dda69eb87c0a37c189b668d0406c05aa8cbbdd56a19bf35  resultados/omniscience.gpt55.jsonl
c19dee5b83b271567bac61c9e4de8e08726a1bc739444493e85edf8f43e22cd9  resultados/omniscience.gpt54.jsonl
ce46be86a95fb48b2e142bdb7fc7c0e063e3ec3888ec0b534ad0a53ea03a636c  resultados/omniscience.grok.jsonl
0ff781c708e5c227bf867b060564c31aa508944e26d7be338daf8d4ad5d1721b  resultados/omniscience.deepseek.jsonl
c67bf6078226b576862c5afbc32ca440a2b11d40b5d89f6ceb553be6f016eac8  resultados/omniscience.kimi.jsonl
4bf96842d6d64026ea3b62f35d0158056af4dc4fe0fa3e36d411922f7cf48492  resultados/simpleqa.house.jsonl
715d4903057616edcc49a722f132b0f6c94646fb295cbf67487db85f7a148967  resultados/simpleqa.gpt55.jsonl
040db462820c1b17dd87e54d589a53cb4ec57195661a60e7726bb1ad02f7ee11  resultados/simpleqa.gpt54.jsonl
3b5bc95969f86c868755b5abd1f672ac3e863af2bd13d523bc74d6d6206d17f0  resultados/simpleqa.grok.jsonl
cbc62179b0b54a9854d1b55f79a81f4422a84423639845c2baa6377a7a1aa0e0  resultados/simpleqa.deepseek.jsonl
d5de43ae0dc1fd27103c64a0b526948d05b7515f78a9aa39d406c1f27312bd68  resultados/simpleqa.kimi.jsonl
e5c399079556bcb992b5356111cd41f0018c6531e5a19e9f967c8768f964a987  resultados/simpleqa_conf.house.rejulgado.jsonl
e12bc360df7519ce9c8f678ca24ca3ca1a8225c773dca9907a3005c29738e01a  resultados/simpleqa_conf.gpt55.jsonl
9550cc0c2e5e18f98676b1c26fc07dac96f7bf0dc82f4a95bb132fc8a92bfe85  resultados/simpleqa_conf.gpt54.jsonl
b6796a56d4d686df5561570423b252893061eacd94f14249670684e8e52a3d9c  resultados/simpleqa_conf.grok.jsonl
d4324f242ecb44a6e32a69fe02afb949e8b1d56fbc1f050cf6e34467673d51cd  resultados/simpleqa_conf.deepseek.jsonl
b85a95a3cb967006dd5423ba65ebf4f2387fea11922c1b21ee3675c27db2468f  resultados/simpleqa_conf.kimi.jsonl
0738386dedd63a42d4746219c821604e0c7f630f39951bd8742aacf71f4da401  resultados/xstest.house.rejulgado.jsonl
e3e9d9bdcc8142dc46770ba8652e63ee9d7b27918555cca40516646f8bbe2327  resultados/xstest.gpt55.jsonl
cbd44ac75f88f98f7e1be8f7cf1971ca5849e0566eca8e19bc76bc8d61052996  resultados/xstest.gpt54.jsonl
40c9bfd7d1b97ddf28fe8763425e55e873fa7e33aba4fa116b03e1e136940b4c  resultados/xstest.grok.jsonl
4b5a2174004230a0b2f179142c574ca4aabdc873f65579bdfde3521b449b265c  resultados/xstest.deepseek.jsonl
abbf947c45ac28f5b0d5515fa7d0e09f17a013d32cd26468619f190251d01b2a  resultados/xstest.kimi.jsonl

Arquivos do pacote técnico que ficaram fora desta versão pública:
  modelos.mjs
  benches/abstention.mjs
  benches/hallulens.mjs
  benches/omniscience.mjs
  benches/orbench.mjs
  benches/orbench_hard_j2.mjs
  benches/orbench_toxic_j2.mjs
  benches/simpleqa.mjs
  benches/simpleqa_conf.mjs
  benches/xstest.mjs
  README.md
  colunas.json
  tabela.json
