SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.20193 · 2026

From Masks to Pixels and Meaning: A New Taxonomy, Benchmark, and Metrics for VLM Image Tampering

Hao Li, Yi Tang, Zhiqiang Shen, Salman Khan, Jiacheng Liu, Xiaohan Zhao, Xinyi Shang, Salwa Al Khatib, Jiacheng Cui, Jing-Hao Xue, Ahmed Elhagry, Sondos Bsharat

arXiv · PDF · Open in the Atlas

Code that ran

2 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 5 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2603.20193.json · badge: /api/ran/2603.20193.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
VILA-Lab/PIXAR community
Papers with Code archive
2 ran · 5 no recorded run
FunctionStatusWhere it lives
dice_loss Ran
ran; our contract draft was wrong, not the code
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("b320489855b18380")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for the function
# inputs: logits from SAM (B,1,H,W or compatible shape)
# batch_size=2, channels=1, height=4, width=3
inputs = torch.randn(2, 1, 4, 3)

# targets: binary/soft masks with values in [0,1]
# same spatial shape as inputs
targets = torch.rand(2, 1, 4, 3)

# num_masks: number of masks (float)
num_masks = 2.0

# eps: epsilon for numerical stability (has default)
eps = 1e-06

ARGS = {
    "inputs": inputs,
    "targets": targets,
    "num_masks": num_masks,
    "eps": eps,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sigmoid_ce_loss Ran
ran; our contract draft was wrong, not the code
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("b9e60254a36a3fb8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn.functional as F

torch.manual_seed(0)

# Create tiny tensors for inputs and targets
# inputs: logits tensor, shape (batch=2, spatial_dims=4)
inputs = torch.randn(2, 4, dtype=torch.float32)

# targets: ground truth binary labels, shape (batch=2, spatial_dims=4)
targets = torch.rand(2, 4, dtype=torch.float32)

# num_masks: number of masks (used for normalization)
num_masks = 2.0

ARGS = {
    "inputs": inputs,
    "targets": targets,
    "num_masks": num_masks,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
LlavaLlamaForCausalLM Not yet run
unverified (harvested, no recorded run)
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("3d4a28c629a82a93")
How it ran

no invocation recorded.

LlavaMetaForCausalLM Not yet run
unverified (harvested, no recorded run)
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("8bf697b4b6b3997f")
How it ran

no invocation recorded.

PIXARForCausalLM Not yet run
unverified (harvested, no recorded run)
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("bd5ae0ddb7768e06")
How it ran

no invocation recorded.

PixarMetaModel Not yet run
unverified (harvested, no recorded run)
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("d151647748d9e916")
How it ran

no invocation recorded.

PixarModel Not yet run
unverified (harvested, no recorded run)
VILA-Lab/PIXAR/model/PIXAR.py
code served (permissive licence) · get_code("09344f737c9d8903")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 7 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Existing tampering detection benchmarks largely rely on object masks, which severely misalign with the true edit signal: many pixels inside a mask are untouched or only trivially modified, while subtle yet consequential edits outside the mask are treated as natural. We reformulate VLM image tampering from coarse region labels to a pixel-grounded, meaning and language-aware task. First, we introduce a taxonomy spanning edit primitives (replace/remove/splice/inpaint/attribute/colorization, etc.) and their semantic class of tampered object, linking low-level changes to high-level understanding. Second, we release a new benchmark with per-pixel tamper maps and paired category supervision to evaluate detection and classification within a unified protocol. Third, we propose a training framework and evaluation metrics that quantify pixel-level correctness with localization to assess confidence or prediction on true edit intensity, and further measure tamper meaning understanding via semanticsaware classification and natural language descriptions for the predicted regions. We also re-evaluate the existing strong segmentation/localization baselines on recent strong tamper detectors and reveal substantial over-and under-scoring using mask-only metrics, and expose failure modes on micro-edits and off-mask changes. Our framework advances the field from masks to pixels, meanings and language descriptions, establishing a rigorous standard for tamper localization, semantic classification and description. Code and benchmark data are available at https://github.com/VILA-Lab/PIXAR.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2603.20193")
get_code_for_paper("2603.20193")
have("2603.20193")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2603.20193.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2603.20193.svg)](https://syntology.ai/paper/2603.20193)

Connect an agent — have() is free.