allenai/OLMo-2-1124-13B-RM warn
Weights only ship in a format that can run code when loaded; its tokenizer differs from its claimed base model; glitch tokens that can silently corrupt ordinary input.
claims base: allenai/OLMo-2-1124-13B-SFT · chat template: present · view on Hugging Face ↗
Scan coverageStatic battery2026-08-25Weights battery2026-08-25Behavioral batterynot rundetails
| Battery | Looks at | Status |
|---|---|---|
| Static battery | Metadata & packaging | complete 2026-08-25 |
| Weights battery | Weights forensics — no GPU, no download | complete 2026-08-25100,280-token embedding scanned · 283 undertrained · lineage consistent · pickle audit clean |
| Behavioral battery | Live-inference differentials | not run |
Ingot runs three batteries against a model. What each one checks →
Findings
Scanned 2026-08-25 · published from a community scan.
medium Pickle-serialized weights, no safetensors
Weights ship only as pickle-based files (pytorch_model-00001-of-00006.bin, pytorch_model-00002-of-00006.bin, pytorch_model-00003-of-00006.bin, …). Loading pickle executes arbitrary code from the file — prefer a safetensors release or load in a sandbox.
How to fix
Convert the weights to safetensors before loading them anywhere that matters.
- Do not load the pickle files in-process — pickle deserialization executes arbitrary code from the file.
- Convert locally in a sandbox: `pip install safetensors` and use `safetensors.torch.save_file` on a state dict loaded with `torch.load(..., weights_only=True)` (refuses most code-execution payloads), or use Hugging Face's `convert.py` space/script.
- Pin the exact revision hash you converted from, and load only your converted safetensors artifact from then on.
medium Vocabulary size differs from claimed parent (100280 vs 100352)
A changed vocab means changed tokenization: strings will split differently than on allenai/OLMo-2-1124-13B-SFT, which can shift behavior on identifiers, codes, and non-English text.
How to fixweight-level
Not patchable: the vocab size mirrors the embedding matrix in the weights. Verify the change was intentional.
- Do not edit `vocab_size` in config.json to "match the parent" — it must equal the embedding table in the shipped weights or the model won't load.
- Diff the tokenizers (`tokenizer.json` / added_tokens) against the parent to see what was added or removed, and test your own identifiers, codes, and non-English text through both.
- If the drift is unexplained by the model card, treat tokenization-sensitive behavior as unvalidated on this model.
medium Undertrained (glitch) token surface in vocabulary
Embedding-norm scan flagged 283 undertrained tokens (norm < 0.3× the vocabulary median of 12.609), including 195 plain-ASCII strings that can appear in ordinary input as identifiers — e.g. "RTCK", "ForCanBeConvertedToF", "AppMethodBeat", "$PostalCodesNL", "Japgolly", "webElementProperties", "(stypy", "adaptiveStyles". 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 Pickle static analysis clean
Opcode-level parse of pytorch_model-00001-of-00006.bin, pytorch_model-00002-of-00006.bin, pytorch_model-00003-of-00006.bin, pytorch_model-00004-of-00006.bin (no code executed) found only standard serialization globals (3 distinct, all torch/collections/numpy). Pickle remains an executable format — this verifies the current bytes, not future uploads; prefer a safetensors release.
info Weights consistent with claimed parent allenai/OLMo-2-1124-13B-SFT
Mean cosine similarity of 64 sampled token-embedding rows against allenai/OLMo-2-1124-13B-SFT is 1.000 — the weights plausibly descend from the declared base (relation: unspecified).
How to fix
Fix or verify the `base_model` declaration so lineage checks can run.
- If you own the repo: correct the `base_model` field in the model card metadata to the real, public parent.
- If you don't: identify the true parent (config architecture + weight shapes narrow it fast) and re-scan with that lineage in mind.
Remediation guidance addresses the documented findings only. It is evidence-driven repair, not a safety certification of the model.
Fingerprint
The durable profile of this model: measured weights-and-metadata facts, rebuilt on every scan and battery run. Updated 2026-08-25.
| architecture | olmo2 · 40 layers · 5120-dim |
| vocabulary | 100,280 tokens |
| license | apache-2.0 |
| serialization | no safetensors pickle |
| chat template | present · sha256:fe689ffbd6a4e2d0 |
| claimed lineage | allenai/OLMo-2-1124-13B-SFT |
| lineage verified | consistent vs allenai/OLMo-2-1124-13B-SFT — embedding-row cosine 1.000 |
| glitch-token surface | 283 undertrained candidates, 195 plain-ASCII |
Full measured fingerprint
| architectures | Olmo2ForSequenceClassification |
| library | transformers |
| pipeline | text-generation |
| repo files | 15 — pickle: pytorch_model-00001-of-00006.bin, pytorch_model-00002-of-00006.bin, pytorch_model-00003-of-00006.bin, pytorch_model-00004-of-00006.bin, pytorch_model-00005-of-00006.bin, pytorch_model-00006-of-00006.bin |
| revision | 2ff069899311 |
| HF snapshot | 619 downloads · 2 likes · updated 2025-01-06 · captured 2026-08-25 |
| pickle audit | pytorch_model-00001-of-00006.bin, pytorch_model-00002-of-00006.bin, pytorch_model-00003-of-00006.bin, pytorch_model-00004-of-00006.bin — 3 standard global(s) |
| embedding tensor | model.embed_tokens.weight · BF16 · 100,280×5120 |
| embedding norms | median 12.6088 · mean 12.3549 |
| lineage check | consistent — cosine 1 over 64 sampled rows vs allenai/OLMo-2-1124-13B-SFT |
| glitch-token samples | "RTCK", "ForCanBeConvertedToF", "AppMethodBeat", "$PostalCodesNL", "Japgolly", "webElementProperties", "(stypy", "adaptiveStyles", "RTLI", "PostalCodesNL", "webElementX", "ForCanBeConverted" |
Battery runs (1)the run trace behind the findings — what each job measured
| battery | status | queued | duration | attempts |
|---|---|---|---|---|
| weights | complete | 2026-08-25 21:47 | 2m | 1 |
weights run 2026-08-25 — measurements
| probes run | glitch-norm-scan, zero-template-token-scan, pickle-static-analysis, lineage-norm-correlation |
| embedding tensor | model.embed_tokens.weight · BF16 · 100,280×5120 |
| glitch surface | 283 undertrained, 195 plain-ASCII |
| lineage check | consistent — cosine 1 over 64 rows vs allenai/OLMo-2-1124-13B-SFT |
Verdict badge
Ship the verdict in your README — it always shows the latest published analysis:
[](https://ingot.tools/models/allenai/OLMo-2-1124-13B-RM)