SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.11297 · NeurIPS · 2021

TokenLearner: What Can 8 Learned Tokens Do for Images and Videos?

Mostafa Dehghani, Michael Ryoo, A Piergiovanni, Anurag Arnab, Anelia Angelova

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 0 failures of our instrument; 0 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/2106.11297.json · badge: /api/ran/2106.11297.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
google-research/scenic canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
rish-16/tokenlearner-pytorch reimplementation (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
SpatialAttention Ran
ran on a synthesized input
rish-16/tokenlearner-pytorch/tokenlearner_pytorch/tokenlearner_pytorch.py
code served (permissive licence) · get_code("292a82cba1f85aa3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned tuple. The image it ran in was not recorded. If your copy returns the same type 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

class SpatialAttention(nn.Module):

    def __init__(self) -> None:
        super().__init__()
        self.conv = nn.Sequential(nn.Conv2d(2, 1, kernel_size=(1, 1), stride=1), nn.BatchNorm2d(1), nn.ReLU())
        self.sgap = nn.AvgPool2d(2)

    def forward(self, x):
        B, H, W, C = x.shape
        x = x.view(B, C, H, W)
        mx = torch.max(x, 1)[0].unsqueeze(1)
        avg = torch.mean(x, 1).unsqueeze(1)
        combined = torch.cat([mx, avg], dim=1)
        fmap = self.conv(combined)
        weight_map = torch.sigmoid(fmap)
        out = (x * weight_map).mean(dim=(-2, -1))
        return (out, x * weight_map)


INIT_ARGS = {}

ARGS = {
    'x': torch.randn(2, 4, 3, 5)
}

if __name__ == '__main__':
    model = SpatialAttention(**INIT_ARGS)
    output = model(**ARGS)
    print("Output shapes:", output[0].shape, output[1].shape)
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
TokenLearner Ran
ran on a synthesized input
rish-16/tokenlearner-pytorch/tokenlearner_pytorch/tokenlearner_pytorch.py
code served (permissive licence) · get_code("ffbb25748372a3c7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, dtype family float, shape (2, 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

class SpatialAttention(nn.Module):

    def __init__(self) -> None:
        super().__init__()
        self.conv = nn.Sequential(nn.Conv2d(2, 1, kernel_size=(1, 1), stride=1), nn.BatchNorm2d(1), nn.ReLU())
        self.sgap = nn.AvgPool2d(2)

    def forward(self, x):
        B, H, W, C = x.shape
        x = x.view(B, C, H, W)
        mx = torch.max(x, 1)[0].unsqueeze(1)
        avg = torch.mean(x, 1).unsqueeze(1)
        combined = torch.cat([mx, avg], dim=1)
        fmap = self.conv(combined)
        weight_map = torch.sigmoid(fmap)
        out = (x * weight_map).mean(dim=(-2, -1))
        return (out, x * weight_map)

class TokenLearner(nn.Module):

    def __init__(self, S) -> None:
        super().__init__()
        self.S = S
        self.tokenizers = nn.ModuleList([SpatialAttention() for _ in range(S)])

    def forward(self, x):
        B, _, _, C = x.shape
        Z = torch.Tensor(B, self.S, C)
        for i in range(self.S):
            Ai, _ = self.tokenizers[i](x)
            Z[:, i, :] = Ai
        return Z


INIT_ARGS = {
    'S': 2
}

ARGS = {
    'x': torch.randn(2, 4, 3, 4)
}

if __name__ == '__main__':
    model = TokenLearner(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output: {output}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 (2, 2, 4) — shape (2, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
TokenLearnerModule Ran
ran on a synthesized input · constructed an object; no result was computed
google-research/scenic/scenic/projects/token_learner/model.py
code served (permissive licence) · get_code("d2eb784d868efe5d")
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 an instance of its own class, TokenLearnerModule: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type 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 math
import jax
import jax.numpy as jnp
import flax.linen as nn

# Initialize JAX random key for reproducibility
key = jax.random.PRNGKey(0)

# Create a small input tensor for testing
# Shape: [bs, h, w, c] = [2, 4, 4, 3]
batch_size = 2
height = 4
width = 4
channels = 3
inputs = jnp.ones((batch_size, height, width, channels))

# Define arguments for TokenLearnerModule
ARGS = {
    'num_tokens': 5,
    'use_sum_pooling': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
flax, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type TokenLearnerModule — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TokenLearnerModule output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
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

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

In this paper, we introduce a novel visual representation learning which relies on a handful of adaptively learned tokens, and which is applicable to both image and video understanding tasks. Instead of relying on hand-designed splitting strategies to obtain visual tokens and processing a large number of densely sampled patches for attention, our approach learns to mine important tokens in visual data. This results in efficiently and effectively finding a few important visual tokens and enables modeling of pairwise attention between such tokens, over a longer temporal horizon for videos, or the spatial content in images. Our experiments demonstrate strong performance on several challenging benchmarks for both image and video recognition tasks. Importantly, due to our tokens being adaptive, we accomplish competitive results at significantly reduced compute amount. We obtain comparable results to the state-of-the-arts on ImageNet while being computationally more efficient. We also confirm the effectiveness of the approach on multiple video datasets, including Kinetics-400, Kinetics-600, Charades, and AViD.

For agents

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

get_harvested_code_for_paper("2106.11297")
get_code_for_paper("2106.11297")
have("2106.11297")

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

curl https://syntology.ai/api/ran/2106.11297.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.11297.svg)](https://syntology.ai/paper/2106.11297)

Connect an agent — have() is free.