SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.12366 · ECCV · 2024

NavGPT-2: Unleashing Navigational Reasoning Capability for Large Vision-Language Models

Qi Wu, Yicong Hong, Zun Wang, Eric Wang, Gengze Zhou

arXiv · PDF · Open in the Atlas

Code that ran

12 of 14 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; 1 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 10 failures of our instrument; 2 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, 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/2407.12366.json · badge: /api/ran/2407.12366.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
GengzeZhou/NavGPT-2 canonical (classifier label)
classifier basis: census
12 ran · 2 no recorded run
FunctionStatusWhere it lives
swish Ran
ran; our contract draft was wrong, not the code
GengzeZhou/NavGPT-2/map_nav_src/models/NavGPT_model.py
code served (permissive licence) · get_code("0f786c407fb1ee4c")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) 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. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-04T22:33:52Z). 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)

# The function swish takes a single parameter 'x' which is used with:
# - torch.sigmoid(x) - requires a tensor
# - x * ... - requires a tensor that supports multiplication
# So x should be a tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
cal_cls Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
GengzeZhou/NavGPT-2/map_nav_src/r2r/eval_utils.py
code served (permissive licence) · get_code("cbef633255b01324")
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 float64, ndim 0, 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 numpy as np

# Create a small shortest_distances matrix
# This is a distance matrix between nodes, so we need it to be square and symmetric
shortest_distances = np.array([
    [0.0, 1.5, 2.0, 3.5],
    [1.5, 0.0, 1.2, 2.1],
    [2.0, 1.2, 0.0, 1.8],
    [3.5, 2.1, 1.8, 0.0]
])

# prediction is a sequence of node indices
prediction = [0, 1, 2, 3]

# reference is a sequence of node indices
reference = [0, 2, 3, 1]

# threshold is already provided as a default parameter (3.0)
threshold = 3.0

ARGS = {
    "shortest_distances": shortest_distances,
    "prediction": prediction,
    "reference": reference,
    "threshold": threshold
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
cal_dtw Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
GengzeZhou/NavGPT-2/map_nav_src/r2r/eval_utils.py
code served (permissive licence) · get_code("7ada68c978baa7c9")
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 dict. 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 numpy as np

# Create shortest_distances: a square matrix representing distances between points
# The function indexes this as shortest_distances[prediction[i]][reference[j]]
# So we need it to be indexable by integers
n_points = 5
shortest_distances = np.random.rand(n_points, n_points)

# Create prediction: a sequence of indices into shortest_distances
prediction = np.array([0, 1, 2, 3], dtype=int)

# Create reference: another sequence of indices into shortest_distances
reference = np.array([0, 2, 4], dtype=int)

# Create success: optional parameter, can be None or a float
success = None

# Create threshold: a float parameter
threshold = 3.0

ARGS = {
    'shortest_distances': shortest_distances,
    'prediction': prediction,
    'reference': reference,
    'success': success,
    'threshold': threshold
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
calc_position_distance Ran
ran, violated the contract we drafted
GengzeZhou/NavGPT-2/map_nav_src/models/graph_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from cshizhe/vln-duet, is pointer only (licence: NONE) · get_code("c362aaed2935ceff")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from cshizhe/vln-duet, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

calculate_vp_rel_pos_fts Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
GengzeZhou/NavGPT-2/map_nav_src/models/graph_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from cshizhe/vln-duet, is pointer only (licence: NONE) · get_code("168e35f8ed1ad0ce")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from cshizhe/vln-duet, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

construct_instrs Ran
ran; our contract draft was wrong, not the code
GengzeZhou/NavGPT-2/map_nav_src/r2r/data_utils.py
code served (permissive licence) · get_code("eb5d2aab5fa48a08")
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 list, 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 os
import json
import tempfile
import shutil

# Create a temporary directory for test data
test_dir = tempfile.mkdtemp()

# Create mock JSON data files for different datasets and splits
mock_data_bert = [
    {
        'id': 'sample_1',
        'path_id': 'path_1',
        'instructions': ['Go forward and turn left', 'Walk to the door'],
        'viewpoint': 'start'
    },
    {
        'id': 'sample_2',
        'path_id': 'path_2',
        'instructions': ['Turn right at the corner', 'Proceed straight'],
        'viewpoint': 'mid'
    }
]

mock_data_reverie = [
    {
        'id': 'reverie_1',
        'instructions': ['Find the red chair', 'Look for objects'],
        'viewpoint': 'start'
    },
    {
        'id': 'reverie_2',
        'instructions': ['Locate the table', 'Move forward'],
        'viewpoint': 'mid'
    }
]

mock_data_rxr = [
    {
        'path_id': 101,
        'instruction_id': 0,
        'instructions': ['Go to the kitchen', 'Find the sink'],
        'viewpoint': 'start'
    },
    {
        'path_id': 102,
        'instruction_id': 1,
        'instructions': ['Walk down the hallway', 'Turn left'],
        'viewpoint': 'mid'
    }
]

# Write mock data files
with open(os.path.join(test_dir, 'R2R_train_enc.json'), 'w') as f:
    json.dump(mock_data_bert, f)

with open(os.path.join(test_dir, 'R2R_val_train_seen_enc.json'), 'w') as f:
    json.dump(mock_data_bert, f)

with open(os.path.join(test_dir, 'REVERIE_train_enc.json'), 'w') as f:
    json.dump(mock_data_reverie, f)

with open(os.path.join(test_dir, 'RXR-EN_train_enc.json'), 'w') as f:
    json.dump(mock_data_rxr, f)

ARGS = {
    'anno_dir': test_dir,
    'dataset': 'R2R',
    'splits': ['train', 'val_train_seen'],
    'tokenizer': 'bert',
    'max_instr_len': 512,
    'is_test': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list 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
extend_neg_masks Ran
ran; our contract draft was wrong, not the code
GengzeZhou/NavGPT-2/map_nav_src/models/ops.py
code served (permissive licence) · get_code("91e1742acf23aa57")
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 4, dtype family float, shape (2, 1, 1, 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

torch.manual_seed(0)

# Construct a small mask tensor of shape (N, L) with binary values
# N=2 (batch size), L=4 (sequence length)
masks = torch.randint(0, 2, (2, 4)).float()

ARGS = {
    "masks": masks,
    "dtype": torch.float,
}
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 4, dtype family float, shape (2, 1, 1, 4) — shape (2, 1, 1, 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
gelu Ran
ran, honoured the contract we drafted
GengzeZhou/NavGPT-2/map_nav_src/models/NavGPT_model.py
pointer only (licence: MIT); the stored copy get_code serves, from tanmay1618/text_to_sql_bert, is pointer only (licence: NONE) · get_code("fdc64f4c72036ae4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from tanmay1618/text_to_sql_bert, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

gen_seq_masks Ran
ran; our contract draft was wrong, not the code
GengzeZhou/NavGPT-2/map_nav_src/models/ops.py
code served (permissive licence) · get_code("d9c76c6dc9976905")
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, 3) 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 a sequence lengths tensor
# The function expects seq_lens to be a 1D tensor with batch_size elements
# It calls .device on it and .unsqueeze(1) on it
seq_lens = torch.tensor([3, 2], dtype=torch.long)  # batch_size=2, values are sequence lengths

# max_len can be None (function will compute it) or an integer
max_len = None

ARGS = {
    "seq_lens": seq_lens,
    "max_len": max_len,
}
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, 3) — shape (2, 3) 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
get_angle_fts Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
GengzeZhou/NavGPT-2/map_nav_src/models/graph_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from cshizhe/vln-duet, is pointer only (licence: NONE) · get_code("dac1993f5ce8e12e")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from cshizhe/vln-duet, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

load_instr_datasets Ran
ran; our contract draft was wrong, not the code
GengzeZhou/NavGPT-2/map_nav_src/r2r/data_utils.py
code served (permissive licence) · get_code("80881fb0e12a6a8e")
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 list, 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 os
import json
import numpy as np
import tempfile

# Create a temporary directory for test data
temp_dir = tempfile.mkdtemp()

# Create test JSON files
test_data_bert = [
    {"instruction": "go to the room", "path": [1, 2, 3]},
    {"instruction": "turn left", "path": [4, 5]},
    {"instruction": "stop", "path": [6]},
]

test_data_xlm = [
    {"instruction": "va a la habitación", "path": [1, 2]},
    {"instruction": "gira a la izquierda", "path": [3, 4, 5]},
]

# Write BERT tokenizer files
with open(os.path.join(temp_dir, 'R4R_train_enc.json'), 'w') as f:
    json.dump(test_data_bert, f)

with open(os.path.join(temp_dir, 'R4R_val_train_seen_enc.json'), 'w') as f:
    json.dump(test_data_bert * 20, f)  # Make it larger so slicing [:50] works

with open(os.path.join(temp_dir, 'R4R_val_unseen_enc.json'), 'w') as f:
    json.dump(test_data_bert * 50, f)  # Large dataset for sampling

# Write XLM tokenizer files
with open(os.path.join(temp_dir, 'R4R_train_enc_xlmr.json'), 'w') as f:
    json.dump(test_data_xlm, f)

with open(os.path.join(temp_dir, 'R4R_val_train_seen_enc_xlmr.json'), 'w') as f:
    json.dump(test_data_xlm * 20, f)

# Create augmented data file for pretraining
augmented_file = os.path.join(temp_dir, 'augmented_data.json')
with open(augmented_file, 'w') as f:
    json.dump(test_data_bert, f)

ARGS = {
    'anno_dir': temp_dir,
    'dataset': 'r4r',
    'splits': ['train', 'val_train_seen'],
    'tokenizer': 'bert',
    'is_test': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list 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
load_obj2vps Ran
ran; our contract draft was wrong, not the code
GengzeZhou/NavGPT-2/map_nav_src/r2r/data_utils.py
code served (permissive licence) · get_code("ab1a50dceb5369de")
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 dict. 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 json
import tempfile
import os

# Create a temporary JSON file with the expected structure
temp_dir = tempfile.mkdtemp()
bbox_file = os.path.join(temp_dir, "bbox.json")

# Create sample bbox data matching the function's expected format
bbox_data = {
    "scan1_vp0": {
        "obj1": {
            "visible_pos": [1, 2, 3]  # non-empty list is truthy
        },
        "obj2": {
            "visible_pos": []  # empty list is falsy
        }
    },
    "scan1_vp1": {
        "obj1": {
            "visible_pos": [4, 5, 6]
        },
        "obj3": {
            "visible_pos": [7, 8, 9]
        }
    },
    "scan2_vp0": {
        "obj4": {
            "visible_pos": [10, 11, 12]
        }
    }
}

# Write the JSON file
with open(bbox_file, 'w') as f:
    json.dump(bbox_data, f)

ARGS = {
    "bbox_file": bbox_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
build_transformer Not yet run
unverified (harvested, no recorded run)
GengzeZhou/NavGPT-2/map_nav_src/models/transformer.py
code served (permissive licence) · get_code("df2919fd1dad1d43")
How it ran

no invocation recorded.

create_transformer_encoder Not yet run
unverified (harvested, no recorded run)
GengzeZhou/NavGPT-2/map_nav_src/models/ops.py
code served (permissive licence) · get_code("c4d81965c22f62f9")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 11 are also present in other papers whose code we hold (11 with at least one paper dated earlier, 2 the same month, 7 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.

FunctionAlso present inEarliest-datedOthers named
swish 206 other papers
181 dated earlier, 21 later, 4 same year, order not determinable · 44 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1805.12114 (2018-05)
Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models
1809.01357 (2018-09)
Zero Shot Learning for Code Education: Rubric Sampling with Deep Learning Inference
and 203 more not named here
gelu 198 other papers
171 dated earlier, 1 the same month, 23 later, 3 same year, order not determinable · 32 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1901.10125 (2019-01)
Glyce: Glyph-vectors for Chinese Character Representations
and 195 more not named here
build_transformer 32 other papers
24 dated earlier, 1 the same month, 6 later, 1 same year, order not determinable
2005.12872 (2020-05)
End-to-End Object Detection with Transformers
2103.12115 (2021-03)
End-to-End Trainable Multi-Instance Pose Estimation with Transformers
2104.04369 (2021-04)
Video-aided Unsupervised Grammar Induction
and 29 more not named here
calc_position_distance 8 other papers
5 dated earlier, 1 later, 2 same year, order not determinable
2202.11742 (2022-02)
Think Global, Act Local: Dual-scale Graph Transformer for Vision-and-Language Navigation
2307.12907 (2023-07)
GridMM: Grid Memory Map for Vision-and-Language Navigation
2308.04758 (2023-08)
Bird's-Eye-View Scene Graph for Vision-Language Navigation
and 5 more not named here
extend_neg_masks 8 other papers
5 dated earlier, 2 later, 1 same year, order not determinable · 4 only as an assembly brick
2303.15796 (2023-03)
KERM: Knowledge Enhanced Reasoning for Vision-and-Language Navigation
2305.03602 (2023-05)
A Dual Semantic-Aware Recurrent Global-Adaptive Network for Vision-and-Language Navigation
2308.04758 (2023-08)
Bird's-Eye-View Scene Graph for Vision-Language Navigation
and 5 more not named here
gen_seq_masks 8 other papers
5 dated earlier, 2 later, 1 same year, order not determinable · 4 only as an assembly brick
2303.15796 (2023-03)
KERM: Knowledge Enhanced Reasoning for Vision-and-Language Navigation
2305.03602 (2023-05)
A Dual Semantic-Aware Recurrent Global-Adaptive Network for Vision-and-Language Navigation
2308.04758 (2023-08)
Bird's-Eye-View Scene Graph for Vision-Language Navigation
and 5 more not named here
get_angle_fts 7 other papers
4 dated earlier, 1 later, 2 same year, order not determinable
2202.11742 (2022-02)
Think Global, Act Local: Dual-scale Graph Transformer for Vision-and-Language Navigation
2308.04758 (2023-08)
Bird's-Eye-View Scene Graph for Vision-Language Navigation
2312.02010 (2023-12)
Towards Learning a Generalist Model for Embodied Navigation
and 4 more not named here
calculate_vp_rel_pos_fts 6 other papers
4 dated earlier, 2 same year, order not determinable
2202.11742 (2022-02)
Think Global, Act Local: Dual-scale Graph Transformer for Vision-and-Language Navigation
2307.12907 (2023-07)
GridMM: Grid Memory Map for Vision-and-Language Navigation
2308.04758 (2023-08)
Bird's-Eye-View Scene Graph for Vision-Language Navigation
and 3 more not named here
cal_cls 2 other papers
1 dated earlier, 1 same year, order not determinable
aaai_28597 (2024 (year only))
NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models
2403.07376 (2024-03)
NavCoT: Boosting LLM-Based Vision-and-Language Navigation via Learning Disentangled Reasoning
cal_dtw 2 other papers
1 dated earlier, 1 same year, order not determinable
aaai_28597 (2024 (year only))
NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models
2403.07376 (2024-03)
NavCoT: Boosting LLM-Based Vision-and-Language Navigation via Learning Disentangled Reasoning
load_instr_datasets 1 other paper
1 dated earlier
2308.04758 (2023-08)
Bird's-Eye-View Scene Graph for Vision-Language Navigation

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

Capitalizing on the remarkable advancements in Large Language Models (LLMs), there is a burgeoning initiative to harness LLMs for instruction following robotic navigation. Such a trend underscores the potential of LLMs to generalize navigational reasoning and diverse language understanding. However, a significant discrepancy in agent performance is observed when integrating LLMs in the Vision-and-Language navigation (VLN) tasks compared to previous downstream specialist models. Furthermore, the inherent capacity of language to interpret and facilitate communication in agent interactions is often underutilized in these integrations. In this work, we strive to bridge the divide between VLN-specialized models and LLM-based navigation paradigms, while maintaining the interpretative prowess of LLMs in generating linguistic navigational reasoning. By aligning visual content in a frozen LLM, we encompass visual observation comprehension for LLMs and exploit a way to incorporate LLMs and navigation policy networks for effective action predictions and navigational reasoning. We demonstrate the data efficiency of the proposed methods and eliminate the gap between LM-based agents and state-of-the-art VLN specialists. The source code is available at https://github.com/GengzeZhou/NavGPT-2.

For agents

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

get_harvested_code_for_paper("2407.12366")
get_code_for_paper("2407.12366")
have("2407.12366")

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

curl https://syntology.ai/api/ran/2407.12366.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.12366.svg)](https://syntology.ai/paper/2407.12366)

Connect an agent — have() is free.