---
license: mit
library_name: transformers
language:
  - en
tags:
  - file-type-detection
  - mime-classification
  - binary-content
  - binary-analysis
  - position-agnostic
  - libmagic
  - forensics
  - packet-inspection
  - byte-level
  - mimelens
pipeline_tag: text-classification
model-index:
  - name: mimelens-001-medium-byte-s1
    results:
      - task:
          type: feature-extraction
          name: MIME-125 classification (libmagic 125-class taxonomy)
        dataset:
          name: magic-frags (4 KB head of 64 KB random chunks, n=4,096)
          type: custom
        metrics:
          - name: top-1 accuracy
            type: accuracy
            value: 0.0000
          - name: macro-F1
            type: f1
            value: 0.0000
          - name: kNN R@1
            type: recall@1
            value: 0.0000
        source:
          name: "MimeLens paper (Bommarito 2026), Appendix A"
          url: https://github.com/mjbommar/mimelens-training
---

# mimelens-001-medium-byte-s1

A 37.76M-backbone-parameter BERT-style encoder for position-agnostic file-content-type detection from binary data. It reads a byte window taken from *any* offset in a file (the first ~1{,}022 tokens of whatever you pass) and produces a 512-dimensional embedding that classifiers map to one of [libmagic](https://github.com/file/file)'s 125 MIME labels. Designed for inputs where you only have a chunk: a forensic-carved fragment, a random disk-block read, a streaming HTTP upload, a single network packet payload.

- **🔗 Model**: [`mjbommar/mimelens-001-medium-byte-s1`](https://huggingface.co/mjbommar/mimelens-001-medium-byte-s1)
- **👥 Family**: [`mjbommar/mimelens-001`](https://huggingface.co/mjbommar/mimelens-001) (36 released cells: 28 parent + 8 short-sequence)
- **📄 Paper**: *MimeLens: Position-Agnostic Content-Type Detection for Binary Fragments* (Bommarito 2026)
- **💻 Training code**: [`mjbommar/mimelens-training`](https://github.com/mjbommar/mimelens-training)
- **📊 Pretraining corpus**: [`mjbommar/binary-30k-tokenized`](https://huggingface.co/datasets/mjbommar/binary-30k-tokenized) plus magic-corpus extracts, packed binaries, a [`glaurung`](https://github.com/mjbommar/glaurung)-sourced binary corpus, and Windows drivers (33 GB stratified; the full corpus is not redistributable)

---

## What MimeLens does

MimeLens classifies file content type from a byte window taken at any offset, not just the header of a complete file.

Existing tools assume whole-file access at a known offset:

- [`libmagic`](https://github.com/file/file) and [Apache Tika](https://tika.apache.org/) match handcrafted magic-byte signatures, almost always anchored at the file head.
- [Magika](https://github.com/google/magika) (Google) is a small (~1 M-parameter) feedforward network over three 512-byte windows (head, middle, tail) of a known-bounded file.
- TrID, PRONOM/Siegfried/DROID similarly require a complete file.

These break down on a fragment. MimeLens is pretrained MLM-only on 1024-token windows sampled *uniformly at random* across files and 64 KB fragments, with no privileged head-of-file position. One checkpoint handles streaming, partial-arrival, mid-file, packet-payload, and forensic-carved inputs uniformly. The trade-off is CPU latency (roughly two orders of magnitude slower than Magika at the medium size; hardware-dependent) in exchange for libmagic's 125-class taxonomy plus position arbitrariness.

The family ships 28 parent cells (3 sizes × 4 vocabs × 2-3 seeds at seq\_len=1024) plus an 8-cell short-sequence extension (medium tier × 4 vocabs × 2 seeds at seq\_len=256). This README documents one of them.

> **This is a short-sequence (`seq_len=256`) cell.** Architecture, optimizer, schedule, and corpus are identical to the parent cube; only the context length differs. For sub-KB deployment regimes (DNS payloads, sub-MTU packets, sub-1 KB forensic fragments), the short-sequence cube runs ~5× faster on CPU and (for BPE cells) ties or beats the seq_len=1024 cube on the magic-files probe-fit benchmark. See paper Appendix B.5.


> **ONNX bundled.** This cell ships `onnx/model_fp32.onnx` + `onnx/model_int8.onnx` (dynamic int8 of MatMul/Gemm) for direct ONNX Runtime inference. See `onnx/README.md` in this repo for input/output shapes and the latency profile.


---

## Overview

- **This cell**: `medium` tier, `byte` input pipeline, seed `1`
- **Backbone**: 37.76M parameters (12 layers, hidden 512, 8 attention heads, head dim 64, RoPE, RMSNorm, no biases, no dropout)
- **Input vocabulary**: `byte`. Raw 256-byte vocabulary plus 5 special tokens (CLS, SEP, PAD, UNK, MASK); id = byte_value + 5. The model reads exactly the first 1,022 bytes that arrive.
- **Output**: 512-dim mean-pooled body-token embedding
- **Label space**: [libmagic](https://github.com/file/file) 125-class MIME taxonomy (full list in paper Appendix)
- **Pretraining**: MLM-only, 30% mask ratio, 33 GB stratified multi-source binary corpus, 22,888 gradient updates, single RTX 4060 Ti, ~4.5 h wall-clock
- **License**: MIT

## Headline benchmarks (this cell)

| Benchmark | Value |
|---|---|
| MIME-125 top-1 (magic-frags, 4 KB head, n=4,096)            | **0.000** |
| MIME-125 macro-F1 (magic-frags, 4 KB head)                  | 0.000 |
| kNN R@1 (magic-frags, 3,147-file gallery / 949 queries)     | 0.000 |
| Δ top-1 under zero-first-16-byte header perturbation        | −0.000 |
| Δ top-1 under zero-first-64-byte header perturbation        | −0.000 |

Full evaluation (within-cube bootstrap CIs, adversarial sweep, calibration, real-network curves, disk-block matrix, baselines against libmagic 5.46 and TrID 2.24) is in the [paper](https://github.com/mjbommar/mimelens-training).

---

## Quick start

This cell ships a 125-class libmagic-MIME classifier head (the paper's LR probe, re-fit on the full magic-files corpus), so `pipeline("text-classification", ...)` works out of the box:

```python
from transformers import pipeline

clf = pipeline("text-classification",
               model="mjbommar/mimelens-001-medium-byte-s1",
               trust_remote_code=True,
               top_k=5)

# The model reads the first ~1,022 tokens of whatever you pass (a prefix of the
# buffer, not the whole window). For whole-file triage, a short head window
# classifies magic-byte / compressed types better than a long one -- see
# "Choosing a window" below.
window = open("path/to/file", "rb").read(4096)
preds  = clf(window.decode("latin-1"))                 # latin-1 is a bijection over bytes
# preds[0] is the list of {label, score} sorted by score:
# [{"label": "image/png", "score": 0.97}, {"label": "image/jpeg", "score": 0.01}, ...]
```

To work with embeddings directly (fit a probe, kNN over a gallery, fine-tune a head):

```python
import torch
from transformers import AutoModel, AutoTokenizer

repo  = "mjbommar/mimelens-001-medium-byte-s1"
model = AutoModel.from_pretrained(repo, trust_remote_code=True).eval()
tok   = AutoTokenizer.from_pretrained(repo)

window = open("path/to/file", "rb").read(4096)
inputs = tok(window.decode("latin-1"), max_length=1024, truncation=True,
             padding="max_length", return_tensors="pt")
with torch.no_grad():
    embedding = model(**inputs).pooler_output         # (1, 512)
```


---

## Choosing a window

The model reads the first ~1{,}022 tokens of whatever you pass — a prefix of the buffer (the first 1{,}022 bytes for this byte cell), not the whole window.

- **Magic-byte / compressed types** (PNG, ZIP, GZIP, JPEG): a **short head window (256 B--1 KB) classifies better than 4 KB**. A long high-entropy body dilutes the header signal within the fixed token budget, and the model returns `application/octet-stream` on a mostly-opaque window — correct behaviour for genuinely high-entropy input, not a bug.
- **Fragments / packets**: you cannot choose the offset, so pass what you have. This is the regime MimeLens is built for.

---

## Recommended deployment regimes

- **Streaming / packet-payload / partial-arrival inputs**: this cell wins by 3.4 pp over the best BPE cell at every cumulative-packet threshold (Appendix D of the paper).
- **kNN retrieval over a chunk store**: tightest within-cube retrieval finding.
- libmagic-taxonomy (125-class) classification from a clean 4 KB chunk (tied with bpe-16k).

---

## Training

This cell is one point of the 3 × 4 × {2,3} factorial cube described in the paper.

- **Corpus** (33 GB, stratified multi-source): [`binary-30k`](https://huggingface.co/datasets/mjbommar/binary-30k-tokenized) (assorted ELF/PE/Mach-O), magic-frags (random 64 KB chunks across libmagic's full corpus), assorted packed/raw binaries, a [`glaurung`](https://github.com/mjbommar/glaurung)-sourced binary corpus, Windows drivers.
- **Position-arbitrary windowing**: 1024-token windows sampled uniformly at random across files and 64 KB fragments. **No privileged "head of file" position.** This is the design choice that makes MimeLens work on streaming / partial / random-offset inputs.
- **Objective**: MLM with 30% mask ratio (BERT replacement schedule: 80% `[MASK]`, 10% random, 10% original); tied input/output embeddings.
- **Pooling**: mean-pool over body tokens for downstream tasks. The BERT-style `cls_pool` linear projection is *not* used: under MLM-only training it receives no gradient and remains byte-identical to its random initialisation across all 28 cube cells (paper §3.4 verifies this; left in the saved weights for architectural completeness only).
- **Optimisation**: AdamW + cosine LR (peak 5e-4, 2,000-step warmup, 10% floor), bf16 mixed precision, gradient clipping at $\|g\|_2 \leq 1$, effective batch 128 at sequence length 1024, 22,888 gradient updates.
- **Hardware**: single RTX 4060 Ti (16 GB), ~4.5 h wall-clock for this cell.

---

## Caveats

- This is one cell of a 28-cell parent cube (36 released cells including the 8-cell short-sequence extension). Within-cube comparisons in the paper carry bootstrap CIs at n=3 seeds; some marginal orderings (byte vs bpe-16k at the largest size) are within seed noise and should be read as ties.
- The training corpus is one 33 GB stratified multi-source binary sample. Results may not transfer to substantially different corpora.
- All numbers are computed on data labelled by a single pipeline (libmagic-pinned). Cross-validation against PRONOM, Siegfried, DROID, or IANA reference files is a documented limitation.
- CPU latency at the `medium` size is ~155× slower than Magika v1.1 on a desktop CPU (hardware-dependent). For sub-millisecond whole-file triage on broad categories, Magika is purpose-built and is the right tool. MimeLens occupies a different point on the deployment surface (position-arbitrary inputs + libmagic's 125-class taxonomy), not a drop-in replacement.
- End-to-end fine-tuning on the production label distribution may shift these numbers and should be evaluated before deployment. The frozen-probe numbers above are not claimed as a lower bound on fine-tuned performance.

---

## Citation

```bibtex
@misc{bommarito2026mimelens,
  title  = {MimeLens: Position-Agnostic Content-Type Detection for Binary Fragments},
  author = {Bommarito II, Michael J.},
  year   = {2026},
  note   = {https://github.com/mjbommar/mimelens-training},
}
```