Data to …… trials
I · Narrative

Belief tracking in deliberately misleading stories

Nemotron 3 Nano 30B

Finding:DEGRADES· grade A(research conclusion)

88% correct (279/317)

Tested up to chain length 10, scramble 1.0, pronoun pressure 5, distractor events 20, witness cutoff 0.9

Pass ratesingle rate, no difficulty dial
88%

279/317 scored trials passed · 95% interval 84%–91%

breaks ≤ 30%holds if interval reaches 95%0%50%100%
Why it slips

The model answers where the object really is, not where the person believes it is.

in 87% of failures (38 failures)

Example failure
Trial result: FAIL (one model answer)
FAIL
lobby
Open this trial →Open the run it came from →
Evidencewhat this finding rests on
  • Ground truthDeterministic simulation: the right answer is computed, not judged
  • StatisticsWilson 95% interval 84%–91%Grade A: strong evidence
  • Sample size317 scored trials in total3 errored trials excluded
  • Fresh-seed confirmationNot run yet
  • ReplicationNot run yet
Raw evidenceruns and trials

Back to the capability map →