allenai/Llama-3.1-8B-Instruct-RM-RB2 warn
Weights only ship in a format that can run code when loaded; the chat template differs from its base model, which changes behavior; its tokenizer differs from its claimed base model. Plus 1 more issue.
claims base: meta-llama/Llama-3.1-8B-Instruct · chat template: present · view on Hugging Face ↗
Scan coverageStatic battery2026-08-21Weights battery2026-08-21Behavioral batterynot rundetails
| Battery | Looks at | Status |
|---|---|---|
| Static battery | Metadata & packaging | complete 2026-08-21 |
| Weights battery | Weights forensics: no GPU, no download | complete 2026-08-21128,264-token embedding scanned · 504 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 2026-08-21 · published from a community scan.
medium Pickle-serialized weights, no safetensors
Weights ship only as pickle-based files (pytorch_model-00001-of-00004.bin, pytorch_model-00002-of-00004.bin, pytorch_model-00003-of-00004.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 Chat template differs from claimed parent
The chat template does not match meta-llama/Llama-3.1-8B-Instruct's. Template drift silently changes model behavior even when weights are identical — 37% of drifted derivatives in our census left it undisclosed. Diff the templates before deploying.
How to fixingot patch
Restore the parent's chat template in `tokenizer_config.json` — a pure metadata fix.
- Run `ingot patch <owner/model>` — the patch manifest carries the parent's template and applies it to a local copy's `tokenizer_config.json`.
- Or fix by hand: copy the `chat_template` value from the parent repo's `tokenizer_config.json` into this model's, and pin your serving stack to that file.
- If the drift was intentional (the author retrained on a new template), confirm that in the model card before "fixing" it — restoring the parent template on retrained weights changes behavior too.
medium Vocabulary size differs from claimed parent (128264 vs 128256)
A changed vocab means changed tokenization: strings will split differently than on meta-llama/Llama-3.1-8B-Instruct, 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 504 undertrained tokens (norm < 0.3× the vocabulary median of 0.685), including 140 plain-ASCII strings that can appear in ordinary input as identifiers — e.g. "ilmektedir", "$PostalCodesNL", "ForCanBeConvertedToF", "TokenNameIdentifier", "CLIIIK", "useRalative", "PostalCodesNL", "_ComCallableWrapper". 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-00004.bin, pytorch_model-00002-of-00004.bin, pytorch_model-00003-of-00004.bin, pytorch_model-00004-of-00004.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 meta-llama/Llama-3.1-8B-Instruct
Mean cosine similarity of 64 sampled token-embedding rows against meta-llama/Llama-3.1-8B-Instruct is 1.000 — the weights plausibly descend from the declared base (relation: unspecified).
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.
Fix it
Some findings are metadata-level and patchable — apply the fixes to your local copy (your weights never leave your machine):
npx @ingotai/scan patch allenai/Llama-3.1-8B-Instruct-RM-RB2
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 2026-08-21.
| architecture | llama · 32 layers · 4096-dim |
| vocabulary | 128,264 tokens |
| license | llama3.1 |
| serialization | no safetensors pickle |
| chat template | present · sha256:ac7498a36a719da6 |
| claimed lineage | meta-llama/Llama-3.1-8B-Instruct |
| lineage verified | consistent vs meta-llama/Llama-3.1-8B-Instruct — embedding-row cosine 1.000 |
| glitch-token surface | 504 undertrained candidates, 140 plain-ASCII |
Full measured fingerprint
| architectures | LlamaForSequenceClassification |
| library | transformers |
| pipeline | text-classification |
| repo files | 11 — pickle: pytorch_model-00001-of-00004.bin, pytorch_model-00002-of-00004.bin, pytorch_model-00003-of-00004.bin, pytorch_model-00004-of-00004.bin |
| revision | 6779f9a30297 |
| HF snapshot | 696 downloads · 1 likes · updated 2025-06-04 · captured 2026-08-21 |
| pickle audit | pytorch_model-00001-of-00004.bin, pytorch_model-00002-of-00004.bin, pytorch_model-00003-of-00004.bin, pytorch_model-00004-of-00004.bin — 3 standard global(s) |
| embedding tensor | model.embed_tokens.weight · BF16 · 128,264×4096 |
| embedding norms | median 0.6849 · mean 0.6713 |
| lineage check | consistent — cosine 1 over 64 sampled rows vs meta-llama/Llama-3.1-8B-Instruct |
| glitch-token samples | "ilmektedir", "$PostalCodesNL", "ForCanBeConvertedToF", "TokenNameIdentifier", "CLIIIK", "useRalative", "PostalCodesNL", "_ComCallableWrapper", "ForCanBeConverted", "krvldkf", "sahuje", "webElementXpaths" |
Battery runs (1)the run trace behind the findings — what each job measured
| battery | status | queued | duration | attempts |
|---|---|---|---|---|
| weights | complete | 2026-08-21 05:13 | 2m | 1 |
weights run 2026-08-21 — measurements
| probes run | glitch-norm-scan, pickle-static-analysis, lineage-norm-correlation |
| embedding tensor | model.embed_tokens.weight · BF16 · 128,264×4096 |
| glitch surface | 504 undertrained, 140 plain-ASCII |
| lineage check | consistent — cosine 1 over 64 rows vs meta-llama/Llama-3.1-8B-Instruct |
Verdict badge
Ship the verdict in your README — it always shows the latest published analysis:
[](https://ingot.tools/models/allenai/Llama-3.1-8B-Instruct-RM-RB2)