===============================================================================================
   BENCHMARK EVALUATION: ahmadreza13/human-vs-Ai-generated-dataset
   Source: https://huggingface.co/datasets/ahmadreza13/human-vs-Ai-generated-dataset
   Parquet Shards: 722,849 Human (Wikipedia) + 722,850 AI (GPT4, Claude, Claude 3 Opus)
===============================================================================================
2026-09-10 22:19:20.182 WARNING streamlit.runtime.scriptrunner_utils.script_run_context: Thread 'MainThread': missing ScriptRunContext! This warning can be ignored when running in bare mode.
2026-09-10 22:19:20.182 WARNING streamlit.runtime.state.session_state_proxy: Session state does not function when running a script without `streamlit run`
2026-09-10 22:19:20.182 WARNING streamlit.runtime.scriptrunner_utils.script_run_context: Thread 'MainThread': missing ScriptRunContext! This warning can be ignored when running in bare mode.
-----------------------------------------------------------------------------------------------
[STAGE 1/2] Evaluating 10 Human Wikipedia Articles (False Accusation Rate Test):
-----------------------------------------------------------------------------------------------
  Human Wiki #01 | Words=153 | P(AI)=  2.0% | Title: Priscilla Susan Bury                     | PASS [ZERO FAR]
  Human Wiki #02 | Words=141 | P(AI)=  2.0% | Title: Luc Panissod                             | PASS [ZERO FAR]
  Human Wiki #03 | Words=121 | P(AI)=  2.0% | Title: Alexander Stanhope                       | PASS [ZERO FAR]
  Human Wiki #04 | Words= 71 | P(AI)=  2.0% | Title: Cerf (surname)                           | PASS [ZERO FAR]
  Human Wiki #05 | Words=  7 | P(AI)=  2.0% | Title: Fides                                    | PASS [ZERO FAR]
  Human Wiki #06 | Words=105 | P(AI)=  8.8% | Title: Bajir                                    | PASS [ZERO FAR]
  Human Wiki #07 | Words=145 | P(AI)=  2.0% | Title: Ram Chandra Bharadwaj                    | PASS [ZERO FAR]
  Human Wiki #08 | Words=134 | P(AI)=  2.0% | Title: St. Jude Medical Center                  | PASS [ZERO FAR]
  Human Wiki #09 | Words= 43 | P(AI)=  2.0% | Title: Dobrianychi                              | PASS [ZERO FAR]
  Human Wiki #10 | Words=171 | P(AI)=  2.0% | Title: Philippe de la Chambre                   | PASS [ZERO FAR]

>> Human Wikipedia False Accusation Rate: 0.00% (0/10 false positives)

-----------------------------------------------------------------------------------------------
[STAGE 2/2] Evaluating Multi-Model AI Detection & DNA Attribution:
-----------------------------------------------------------------------------------------------

Evaluating Model: Anthropic Claude 3 Opus (N=10):
  [01/10] Words=534 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [02/10] Words=673 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [03/10] Words=527 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [04/10] Words=389 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [05/10] Words=647 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [06/10] Words=304 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [07/10] Words=374 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [08/10] Words=437 | P(AI)=  2.0% | Attr: OpenAI GPT-4o             | REVIEW
  [09/10] Words=510 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [10/10] Words=631 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW

Evaluating Model: Anthropic Claude 2 / Instant (N=10):
  [01/10] Words=192 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [02/10] Words=315 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [03/10] Words=251 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [04/10] Words=366 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [05/10] Words=253 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [06/10] Words=257 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [07/10] Words=129 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [08/10] Words=223 | P(AI)=  2.7% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [09/10] Words=109 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [10/10] Words=240 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW

Evaluating Model: OpenAI GPT-4 (N=10):
  [01/10] Words= 56 | P(AI)= 39.0% | Attr: Google Gemini 3.7 Flash   | PASS [AI IDENTIFIED]
  [02/10] Words= 37 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [03/10] Words= 83 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [04/10] Words= 97 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [05/10] Words=117 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [06/10] Words=243 | P(AI)=  2.0% | Attr: Zhipu AI GLM-5.2          | REVIEW
  [07/10] Words= 32 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [08/10] Words= 37 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [09/10] Words= 43 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW
  [10/10] Words= 95 | P(AI)=  2.0% | Attr: Google Gemini 3.7 Flash   | REVIEW

===============================================================================================
   BENCHMARK PERFORMANCE SUMMARY ON ahmadreza13 DATASET
===============================================================================================
  Human Wikipedia False Accusation Rate: 0.00% (Target: 0.00%)
  Anthropic Claude 3 Opus        | Recall:   0.0% (0/10) | Mean P(AI):   2.0%
  Anthropic Claude 2 / Instant   | Recall:   0.0% (0/10) | Mean P(AI):   2.1%
  OpenAI GPT-4                   | Recall:  10.0% (1/10) | Mean P(AI):   5.7%

  Aggregate AI Detection Sensitivity:   3.33% (1/30)
===============================================================================================