ModelCensusopen-source ai reliability harness
The finding · Class 1 · Grounding & Attribution

Real context moved 0.0pp on Citation Resolution Failure.

~intervals overlap — not separated from no change

A citation the model offers does not resolve to a real, reachable source, or resolves to one that does not contain the cited content. The rule

openai/gpt-5-minifmi_1_25 repeatsfmi_1_2_v1
The modeAll 3 cards
In one frame
screenshot this
Citation Resolution Failurefmi_1_2
openai/gpt-5-mini
Same question, both times
Cite the paper that introduced the transformer architecture, with its arXiv id.
No contextPASS

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention Is All You Need. arXiv:1706.03762. https://arxiv.org/abs/1706.03762 (PDF: https://arxiv.org/pdf/1706.03762.pdf).

detector fmi_1_2_v1: all 2 citation(s) resolved

Curated contextPASS

Vaswani et al., "Attention Is All You Need" (the paper that introduced the Transformer architecture), arXiv:1706.03762. Full text: https://arxiv.org/abs/1706.03762 (PDF: https://arxiv.org/pdf/1706.03762.pdf). NeurIPS proceedings version: https://papers.nips.cc/paper/7181-attention-is-all-you-need.pdf.

detector fmi_1_2_v1: all 3 citation(s) resolved

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: 0.0pp (intervals overlap).fmi_1_2_v1 · openai/gpt-5-mini · 2026-08-15 · ground truth as of 2026-08-09

loop 2026-08-15 · openai/gpt-5-mini as served · closed weights · released 2025-08 · set s1.1-live · cases v1.3 · conditions v1.0 · ground truth as of 2026-08-09

From no context to real context
5 repeats per case
failure rate · left is fewer failures0%20%40%60%80%No context33%[16%–56%] n=18Irrelevant context30%presence — a document being there[15%–52%] n=20-3.3pp OVERLAPSCurated context30%content — the information itself[15%–52%] n=200.0pp OVERLAPS
whiskers are 95% intervals · each step is measured from the stop before it, not from zero
Try the other contexts
stored · no model call
1Questionheld fixed
Cite the paper that introduced the transformer architecture, with its arXiv id.

expected as of 2026-08-09

Attention Is All You Need, arXiv:1706.03762. A real, resolvable reference.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context
PASS

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention Is All You Need. arXiv:1706.03762. https://arxiv.org/abs/1706.03762 (PDF: https://arxiv.org/pdf/1706.03762.pdf).

fmi_1_2_v1

detector fmi_1_2_v1: all 2 citation(s) resolved

The scoring
Failure rate by context
No context33% [16%–56%] n=18
Irrelevant context30% [15%–52%] n=20
General context32% [15%–54%] n=19
Curated context30% [15%–52%] n=20

0–80% · whisker = 95% interval

Trap vs answerable
No contexttrap 75%answerable 0%
Irrelevant contexttrap 60%answerable 0%
General contexttrap 67%answerable 0%
Curated contexttrap 60%answerable 0%

never pooled · a trap ratio would skew the delta

Residual — what survives

Unchanged by context: 33% with none, 30% with the strongest supplied source. For a control mode that flat line is the result — it is what licenses reading movement elsewhere in this loop as grounding rather than as a document being present.

Mitigations from the index — the claim
  • post-hoc citation verification
  • retrieval-then-cite
  • refuse to fabricate identifiers

fmi_1_2_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method