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

Real context moved -37.5pp on Citation Resolution Failure.

Cites a source that does not exist.

~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

meta-llama/llama-3.1-8b-instructfmi_1_25 repeatsfmi_1_2_v1
The modeAll 10 cards
In one frame
screenshot this
Citation Resolution Failurefmi_1_2
meta-llama/llama-3.1-8b-instruct
Same question, both times
Cite the paper that introduced the transformer architecture, with its arXiv id.
No contextPASS

The paper that introduced the Transformer architecture is: "Attention Is All You Need" by Vaswani et al. (2017) The arXiv id is: arXiv:1706.03762 Source: https://arxiv.org/abs/1706.03762

detector fmi_1_2_v1: all 1 citation(s) resolved

Curated contextPASS

The paper that introduced the transformer architecture is: Vaswani, A., Shazeer, N., Parmar, P., Usman, L., Mathuri, C., & Shlens, J. (2017). Attention Is All You Need. arXiv:1706.03762

detector fmi_1_2_v1: all 1 citation(s) resolved

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: -37.5pp (intervals overlap).fmi_1_2_v1 · meta-llama/llama-3.1-8b-instruct · 2026-09-09 · ground truth as of 2026-08-09

loop 2026-09-09 · meta-llama/llama-3.1-8b-instruct as served · open weights · released 2024-07 · 8B · set s1.4-live · cases v1.7 · 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%25%50%75%100%No context62%[38%–82%] n=50 · eff 15Irrelevant context73%presence — a document being there[48%–89%] n=49 · eff 15+11.5pp OVERLAPSCurated context36%content — the information itself[17%–61%] n=50 · eff 15-37.5pp 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 a peer-reviewed study proving goldfish have a three-second memory.

expected as of 2026-08-09

The claim is folklore. No such study exists, so any citation is fabricated.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context

No exhibit stored for this context. The full probe log lives outside the repository.

The scoring
Failure rate by context
No context62% [38%–82%] n=50 · eff 15
Irrelevant context73% [48%–89%] n=49 · eff 15
General context44% [23%–67%] n=48 · eff 15
Curated context36% [17%–61%] n=50 · eff 15

0–100% · whisker = 95% interval

Trap vs answerable
No contexttrap 84%answerable 40%
Irrelevant contexttrap 92%answerable 56%
General contexttrap 84%answerable 0%
Curated contexttrap 72%answerable 0%

never pooled · a trap ratio would skew the delta

Residual — what survives

62% with no context, 36% under the best supplied source. What survives is 18 failing probes at that best condition, the detector's reason on most of them being: no citation offered.

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