danish-foundation-models/DFM-Mimir warn
Glitch tokens that can silently corrupt ordinary input.
chat template: present · view on Hugging Face ↗
Scan coverageStatic battery2026-08-27Weights battery2026-08-25Behavioral batterycompletedetails
| Battery | Looks at | Status |
|---|---|---|
| Static battery | Metadata & packaging | complete 2026-08-27 |
| Weights battery | Weights forensics: no GPU, no download | complete 2026-08-25262,144-token embedding scanned · 38050 undertrained |
| Behavioral battery | Live-inference differentials | completefull differential battery (curated) |
Ingot runs three batteries against a model. What each one checks →
Findings
Scanned 2026-08-27 · published from a community scan.
medium Undertrained (glitch) token surface in vocabulary
Embedding-norm scan flagged 38050 undertrained tokens (norm < 0.3× the vocabulary median of 0.705), including 11917 plain-ASCII strings that can appear in ordinary input as identifiers — e.g. "<unused4598>", "<unused147>", "btnbrowser", "<unused873>", "<unused2489>", "HistoryMarks", "<unused1874>", "<unused130>". In models where this class was tested behaviorally, such tokens silently rewrote user input into confident, schema-valid, wrong output. These are candidates from the weights alone; behavioral confirmation requires the behavioral battery.
How to fixruntime guardweight-level
Keep the affected token strings out of the model's input — the scan-derived runtime guard carries this model's exact blocklist.
- Fetch this model's guard artifact (`/api/v1/guard/<owner>/<model>`): the confirmed corrupting tokens and the low-norm candidate list, derived from the published scan.
- Screen inbound text with it (the `@ingotai/guard` package is a reference implementation) and route flagged records to a different model or human review — verbatim-copy tasks on flagged strings are the failure mode.
- The underlying cause is undertrained embeddings in the weights; a true fix is weight-level (continued pretraining on the affected tokens) — that is not a patch, it's a training job.
info Glitch-token echo probe clean
The model repeated 13/16 low-norm candidate tokens verbatim (controls 5/8). No behavioral glitch-token differential at this threshold.
Remediation guidance addresses the documented findings only. It is evidence-driven repair, not a safety certification of the model.
Check every checkpoint before it ships
Use the web app, API, CLI, or CI gate to scan candidate checkpoints and catch model drift before deployment. Public-model scans publish to the open database; paid plans add the volume needed for continuous checks.
Fingerprint
The durable profile of this model: measured weights-and-metadata facts, rebuilt on every scan and battery run. Updated 2026-08-25.
| architecture | hrm_text · 16 layers · 1536-dim |
| parameters | 1786.8M |
| vocabulary | 262,144 tokens |
| license | apache-2.0 |
| serialization | safetensors |
| chat template | present (chat_template.jinja) · sha256:33204f1acb5bd000 |
| glitch-token surface | 38,050 undertrained candidates, 11,917 plain-ASCII |
Full measured fingerprint
| architectures | HrmTextForCausalLM |
| library | transformers |
| pipeline | text-generation |
| repo files | 10 |
| revision | 2844f0178e69 |
| HF snapshot | 5.1k downloads · 50 likes · updated 2026-08-20 · captured 2026-08-25 |
| embedding tensor | model.embed_tokens.weight · BF16 · 262,144×1536 |
| embedding norms | median 0.7048 · mean 0.6899 |
| lineage check | no claimed base model |
| glitch-token samples | "<unused4598>", "<unused147>", "btnbrowser", "<unused873>", "<unused2489>", "HistoryMarks", "<unused1874>", "<unused130>", "<unused289>", "<unused5881>", "<unused3110>", "<unused683>" |
Battery runs (3)the run trace behind the findings — what each job measured
| battery | status | queued | duration | attempts |
|---|---|---|---|---|
| gpu | complete | 2026-08-27 04:31 | 3m | 1 |
| gpu | complete | 2026-08-25 21:47 | 2m | 1 |
| weights | complete | 2026-08-25 19:00 | 33s | 1 |
gpu run 2026-08-27 — measurements
| probes run | glitch |
gpu run 2026-08-25 — measurements
| probes run | glitch |
weights run 2026-08-25 — measurements
| probes run | glitch-norm-scan, zero-template-token-scan, lineage-norm-correlation |
| embedding tensor | model.embed_tokens.weight · BF16 · 262,144×1536 |
| glitch surface | 38,050 undertrained, 11,917 plain-ASCII |
| lineage check | not checked (no claimed base model) |
Verdict badge
Ship the verdict in your README — it always shows the latest published analysis:
[](https://ingot.tools/models/danish-foundation-models/DFM-Mimir)