SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.03214 · 2024

LeGrad: An Explainability Method for Vision Transformers via Feature Formation Sensitivity

arXiv · PDF · Open in the Atlas

Code that ran

2 of 3 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; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2404.03214.json · badge: /api/ran/2404.03214.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
walbouss/legrad canonical (classifier label)
classifier basis: census
2 ran · 1 no recorded run
FunctionStatusWhere it lives
hooked_attention_forward Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
walbouss/legrad/legrad/utils.py
code served (permissive licence) · get_code("7ff0fc26ab9ea663")
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 3, dtype family float, shape (5, 2, 4) 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
import torch.nn as nn
from torch.nn import functional as F

torch.manual_seed(0)

class MockMultiheadAttention(nn.Module):
    def __init__(self, embed_dim, num_heads):
        super().__init__()
        self.embed_dim = embed_dim
        self.num_heads = num_heads
        self.head_dim = embed_dim // num_heads
        
        # Create in_proj_weight and in_proj_bias for q, k, v projections
        self.in_proj_weight = nn.Parameter(torch.randn(3 * embed_dim, embed_dim))
        self.in_proj_bias = nn.Parameter(torch.randn(3 * embed_dim))
        
        # Output projection
        self.out_proj = nn.Linear(embed_dim, embed_dim)
        
        # Storage for attention map (used in hooked_attention_forward)
        self.attention_map = None

# Set up parameters
L = 5  # sequence length
N = 2  # batch size
C = 4  # embedding dimension
num_heads = 2

# Create the mock attention module
module = MockMultiheadAttention(embed_dim=C, num_heads=num_heads)

# Create input tensors
x = torch.randn(L, N, C)
x_k = torch.randn(L, N, C)
x_v = torch.randn(L, N, C)

# Create optional attention mask (upper triangular causal mask)
attn_mask = torch.triu(torch.ones(L, L, dtype=torch.bool), diagonal=1)

ARGS = {
    "self": module,
    "x": x,
    "x_k": x_k,
    "x_v": x_v,
    "attn_mask": attn_mask,
    "need_weights": False,
}
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 3, dtype family float, shape (5, 2, 4) — shape (5, 2, 4) 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
hooked_resblock_forward Ran
ran; our contract draft was wrong, not the code
walbouss/legrad/legrad/utils.py
code served (permissive licence) · get_code("6474f0ede636d6e3")
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 2, dtype family float, shape (2, 4) 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
import torch.nn as nn

torch.manual_seed(0)

# Build minimal attention module that accepts q_x, k_x, v_x, attn_mask
class MinimalAttention(nn.Module):
    def forward(self, q_x, k_x=None, v_x=None, attn_mask=None):
        # Simple passthrough that returns a tensor of same shape as q_x
        return torch.zeros_like(q_x)

# Build minimal MLP module
class MinimalMLP(nn.Module):
    def forward(self, x):
        return torch.zeros_like(x)

# Build the hooked resblock module with required attributes
class HookedResBlock(nn.Module):
    def __init__(self):
        super().__init__()
        self.ln_1 = nn.LayerNorm(4)  # for q_x
        self.ln_1_kv = nn.LayerNorm(4)  # for k_x and v_x
        self.ln_2 = nn.LayerNorm(4)
        self.attention = MinimalAttention()
        self.mlp = MinimalMLP()
        # LayerScale modules (ls_1, ls_2)
        self.ls_1 = nn.Identity()  # Simple identity for layer scale
        self.ls_2 = nn.Identity()  # Simple identity for layer scale
        # These will be set during forward pass
        self.feat_post_attn = None
        self.feat_post_mlp = None

# Create instance
resblock = HookedResBlock()

# Create input tensors (batch=2, dim=4)
q_x = torch.randn(2, 4)
k_x = torch.randn(2, 4)
v_x = torch.randn(2, 4)
attn_mask = None  # Optional, can be None

ARGS = {
    'self': resblock,
    'q_x': q_x,
    'k_x': k_x,
    'v_x': v_x,
    'attn_mask': attn_mask
}
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 2, dtype family float, shape (2, 4) — shape (2, 4) 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
hooked_torch_multi_head_attention_forward Not yet run
unverified (harvested, no recorded run)
walbouss/legrad/legrad/utils.py
code served (permissive licence) · get_code("069f9c77922d657e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 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 3 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

Vision Transformers (ViTs), with their ability to model long-range dependencies through self-attention mechanisms, have become a standard architecture in computer vision. However, the interpretability of these models remains a challenge. To address this, we propose LeGrad, an explainability method specifically designed for ViTs. LeGrad computes the gradient with respect to the attention maps of ViT layers, considering the gradient itself as the explainability signal. We aggregate the signal over all layers, combining the activations of the last as well as intermediate tokens to produce the merged explainability map. This makes LeGrad a conceptually simple and an easy-to-implement tool for enhancing the transparency of ViTs. We evaluate LeGrad in challenging segmentation, perturbation, and open-vocabulary settings, showcasing its versatility compared to other SotA explainability methods demonstrating its superior spatial fidelity and robustness to perturbations. A demo and the code is available at https://github.com/WalBouss/LeGrad.

For agents

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

get_harvested_code_for_paper("2404.03214")
get_code_for_paper("2404.03214")
have("2404.03214")

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

curl https://syntology.ai/api/ran/2404.03214.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.03214.svg)](https://syntology.ai/paper/2404.03214)

Connect an agent — have() is free.