SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2412.11673 · NeurIPS · 2025

DINO-Foresight: Looking into the Future with DINO

Valeo Ai, Spyros Gidaris, Efstathios Karypidis, Ioannis Kakogeorgiou, Nikos Komodakis, Athena Archimedes, Greece Research Center

arXiv · PDF · Open in the Atlas

Code that ran

3 of 10 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; 2 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 0 failures of our instrument; 7 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/2412.11673.json · badge: /api/ran/2412.11673.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
sta8is/dino-foresight marked official in Papers with Code
Papers with Code archive
3 ran · 7 no recorded run
FunctionStatusWhere it lives
AddBroadcastPosEmbed Ran
ran on a synthesized input
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("a5d1c8b433623f9d")
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 4, dtype family float, shape (1, 4, 3, 8) 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 AddBroadcastPosEmbed(nn.Module):

    def __init__(self, shape, embd_dim, dim=-1):
        super().__init__()
        assert dim in [-1, 1]
        self.shape = shape
        self.n_dim = n_dim = len(shape)
        self.embd_dim = embd_dim
        self.dim = dim
        assert embd_dim % n_dim == 0, f'{embd_dim} % {n_dim} != 0'
        self.emb = nn.ParameterDict({f'd_{i}': nn.init.trunc_normal_(nn.Parameter(torch.randn(self.shape[i], embd_dim // n_dim)), 0.0, 0.02) if dim == -1 else nn.init.trunc_normal_(torch.randn(embd_dim // n_dim, self.shape[i]), 0.0, 0.02) for i in range(n_dim)})

    def forward(self, x, decode_step=None, decode_idx=None):
        embs = []
        for i in range(self.n_dim):
            e = self.emb[f'd_{i}']
            if self.dim == -1:
                e = e.view(1, *(1,) * i, self.shape[i], *(1,) * (self.n_dim - i - 1), -1)
                e = e.expand(1, *self.shape, -1)
            else:
                e = e.view(1, -1, *(1,) * i, self.shape[i], *(1,) * (self.n_dim - i - 1))
                e = e.expand(1, -1, *self.shape)
            embs.append(e)
        embs = torch.cat(embs, dim=self.dim)
        return x + embs


INIT_ARGS = {
    'shape': (4, 3),
    'embd_dim': 8,
    'dim': -1
}

ARGS = {
    'x': torch.randn(1, 4, 3, 8),
    'decode_step': None,
    'decode_idx': None
}

if __name__ == '__main__':
    model = AddBroadcastPosEmbed(**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 4, dtype family float, shape (1, 4, 3, 8) — shape (1, 4, 3, 8) 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)
compute_depth_metrics Ran
ran, honoured the contract we drafted
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("9647212f40af1789")
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 tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

torch.manual_seed(0)

class MetricMock:
    """Mock metric object that has a compute() method."""
    def __init__(self, value):
        self.value = value
    
    def compute(self):
        return self.value

# Create mock metric objects with small tensor/scalar values
d1_m = MetricMock(torch.tensor(0.85))
d2_m = MetricMock(torch.tensor(0.92))
d3_m = MetricMock(torch.tensor(0.78))
abs_rel_m = MetricMock(torch.tensor(0.12))
rmse_m = MetricMock(torch.tensor(0.45))
log_10_m = MetricMock(torch.tensor(0.08))
rmse_log_m = MetricMock(torch.tensor(0.15))
silog_m = MetricMock(torch.tensor(0.22))
sq_rel_m = MetricMock(torch.tensor(0.18))

ARGS = {
    'd1_m': d1_m,
    'd2_m': d2_m,
    'd3_m': d3_m,
    'abs_rel_m': abs_rel_m,
    'rmse_m': rmse_m,
    'log_10_m': log_10_m,
    'rmse_log_m': rmse_log_m,
    'silog_m': silog_m,
    'sq_rel_m': sq_rel_m,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
compute_normal_metrics Ran
ran, honoured the contract we drafted
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("1ddac064686a8021")
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 tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

torch.manual_seed(0)

class MockMetric:
    def __init__(self, value):
        self.value = value
    
    def compute(self):
        return self.value

mean_ae_m = MockMetric(torch.tensor(0.5))
median_ae_m = MockMetric(torch.tensor(0.45))
rmse_m = MockMetric(torch.tensor(0.65))
a1_m = MockMetric(torch.tensor(0.8))
a2_m = MockMetric(torch.tensor(0.9))
a3_m = MockMetric(torch.tensor(0.95))
a4_m = MockMetric(torch.tensor(0.97))
a5_m = MockMetric(torch.tensor(0.98))

ARGS = {
    'mean_ae_m': mean_ae_m,
    'median_ae_m': median_ae_m,
    'rmse_m': rmse_m,
    'a1_m': a1_m,
    'a2_m': a2_m,
    'a3_m': a3_m,
    'a4_m': a4_m,
    'a5_m': a5_m,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
DPTHead Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("4107e3b3305e9452")
How it ran

no invocation recorded.

Dino_f Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("ad89576b6df002d9")
How it ran

no invocation recorded.

MaskTransformer Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("0ec469a5039f384d")
How it ran

no invocation recorded.

reset_depth_metrics Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("0f7a80d0fe355d88")
How it ran

no invocation recorded.

reset_normal_metrics Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("812540a84a99f16c")
How it ran

no invocation recorded.

update_depth_metrics Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("cc7f93b1ea98f3b6")
How it ran

no invocation recorded.

update_normal_metrics Not yet run
unverified (harvested, no recorded run)
sta8is/dino-foresight/src/dino_f.py
code served (permissive licence) · get_code("fdbc405da9b53f2c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 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 10 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

Predicting future dynamics is crucial for applications like autonomous driving and robotics, where understanding the environment is key. Existing pixel-level methods are computationally expensive and often focus on irrelevant details. To address these challenges, we introduce DINO-Foresight, a novel framework that operates in the semantic feature space of pretrained Vision Foundation Models (VFMs). Our approach trains a masked feature transformer in a self-supervised manner to predict the evolution of VFM features over time. By forecasting these features, we can apply off-the-shelf, task-specific heads for various scene understanding tasks. In this framework, VFM features are treated as a latent space, to which different heads attach to perform specific tasks for future-frame analysis. Extensive experiments show the very strong performance, robustness and scalability of our framework. Project page and code at https://dino-foresight.github.io/. 39th Conference on Neural Information Processing Systems (NeurIPS 2025).

For agents

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

get_harvested_code_for_paper("2412.11673")
get_code_for_paper("2412.11673")
have("2412.11673")

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

curl https://syntology.ai/api/ran/2412.11673.json

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

[![Syntology run record](https://syntology.ai/api/ran/2412.11673.svg)](https://syntology.ai/paper/2412.11673)

Connect an agent — have() is free.