Benchmark Almanac

PsyCoMark: Conspiracy Detection

NJUST_KMG 0.89 tops its leaderboard
Not in model cards

Models decide whether a Reddit comment expresses a conspiracy belief. The organizers score predictions against human labels using weighted F1.

The organizers report 52 teams on CodaBench and publish scores for 28 qualifying teams. The winning team used GPT-5.2. The organizers report 0.89 weighted F1.

Released
March 2025
Kind
Reading comprehension
Level
School
Scored
Checked automatically
Answers
Prediction
Tasks
Public, plus a private test set
Measures
Capability
Subject
Psychology
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
A host
28
Entries in its paper, 0 models we track

NJUST_KMG tops the leaderboard with 0.89

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

  1. 1 NJUST_KMG 0.89
  2. 2 dangphuduy 0.78
  3. 3 mdok-style 0.78
  4. 4 VARH-AI 0.78
  5. 5 UMUTeam 0.77
  6. 6 CuriosAI 0.76
  7. 7 HU 0.76
  8. 8 NUST PsyAI 0.76
  9. 9 AILS-NTUA 0.75
  10. 10 CSECU-DSG 0.75
  11. 11 davidinfotec 0.75
  12. 12 LATE-iimas 0.75
  13. 13 psy_detectives 0.75
  14. 14 TruthGradient 0.75
  15. 15 Unibuc-NLP 0.74
  16. 16 wangkongqiang 0.74
  17. 17 YNU-HPCC 0.74
  18. 18 CCNU 0.73
  19. 19 Hidetsune 0.73
  20. 20 Macaroni 0.73
  21. 21 Team A 0.73
  22. 22 UCSC NLP 0.73
  23. 23 zhangpeng 0.73
  24. 24 TTLab 0.72
  25. 25 AGAI 0.70
  26. 26 CUET_320 0.59
  27. 27 Jia 0.37
  28. 28 GUNLP 0.34

Source: The PsyCoMark: Conspiracy Detection paper

#EntryScoreSettingSubmitted byDateSource
1 NJUST_KMG 0.89 NJUST_KMG Official leaderboard
2 dangphuduy 0.78 dangphuduy Official leaderboard
3 mdok-style 0.78 mdok-style Official leaderboard
4 VARH-AI 0.78 VARH-AI Official leaderboard
5 UMUTeam 0.77 UMUTeam Official leaderboard
6 CuriosAI 0.76 CuriosAI Official leaderboard
7 HU 0.76 HU Official leaderboard
8 NUST PsyAI 0.76 NUST PsyAI Official leaderboard
9 AILS-NTUA 0.75 AILS-NTUA Official leaderboard
10 CSECU-DSG 0.75 CSECU-DSG Official leaderboard
11 davidinfotec 0.75 davidinfotec Official leaderboard
12 LATE-iimas 0.75 LATE-iimas Official leaderboard
13 psy_detectives 0.75 psy_detectives Official leaderboard
14 TruthGradient 0.75 TruthGradient Official leaderboard
15 Unibuc-NLP 0.74 Unibuc-NLP Official leaderboard
16 wangkongqiang 0.74 wangkongqiang Official leaderboard
17 YNU-HPCC 0.74 YNU-HPCC Official leaderboard
18 CCNU 0.73 CCNU Official leaderboard
19 Hidetsune 0.73 Hidetsune Official leaderboard
20 Macaroni 0.73 Macaroni Official leaderboard
21 Team A 0.73 Team A Official leaderboard
22 UCSC NLP 0.73 UCSC NLP Official leaderboard
23 zhangpeng 0.73 zhangpeng Official leaderboard
24 TTLab 0.72 TTLab Official leaderboard
25 AGAI 0.70 AGAI Official leaderboard
26 CUET_320 0.59 CUET_320 Official leaderboard
27 Jia 0.37 Jia Official leaderboard
28 GUNLP 0.34 GUNLP Official leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Knowledge

All 332 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy. Seed1.8 92.3%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%

Data audit

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the benchmark’s paper (28 entries).
  • Pass Models scored 28 entries on its leaderboard, none a model we track.
  • Note Newest score No scores yet.
  • Note Frontier models No models scored yet.
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on codabench.org: 28 entries at the last read, none a model we track.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check