SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.07707 · ICCV · 2025

PRVQL: Progressive Knowledge-guided Refinement for Robust Egocentric Visual Query Localization

Yan Huang, James Liang, Yapeng Tian, Bing Fan, Heng Fan, Yunhe Feng, Yuewei Lin

arXiv · PDF · Open in the Atlas

Code that ran

11 of 13 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; 3 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 0 failures of our instrument; 2 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/2502.07707.json · badge: /api/ran/2502.07707.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
fb-reps/PRVQL canonical (classifier label)
classifier basis not recorded on the link
11 ran · 2 no recorded run
FunctionStatusWhere it lives
extract_sub_modules_state_dict Ran
ran on a synthesized input
fb-reps/PRVQL/model/corr_clip_spatial_transformer2_anchor_2heads_hnm.py
code served (permissive licence) · get_code("e58720b43049fef2")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create a checkpoint dictionary with the expected structure
# The function filters keys by prefix and removes the prefix from the keys
checkpoint = {
    'clipMatcher_base.backbone.layer1.weight': torch.randn(4, 3),
    'clipMatcher_base.backbone.layer1.bias': torch.randn(4),
    'clipMatcher_base.backbone.layer2.weight': torch.randn(5, 4),
    'clipMatcher_base.backbone.layer2.bias': torch.randn(5),
    'other_module.weight': torch.randn(3, 3),  # This should be filtered out
    'clipMatcher_base.backbone.conv.weight': torch.randn(2, 3, 3, 3),
}

ARGS = {
    'checkpoint': checkpoint,
    'prefix': 'clipMatcher_base.backbone.'
}
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 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
format_predictions Ran
ran on a synthesized input
fb-reps/PRVQL/inference_results.py
code served (permissive licence) · get_code("fe7c2f09c012c022")
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 torch

# Small stand-in class for predicted_rts values
class PredictedRT:
    def to_json(self):
        return {'bboxes': [[0.1, 0.2, 0.3, 0.4]], 'score': 0.95}

# Build annotations structure
annotations = {
    'version': '1.0',
    'videos': [
        {
            'video_uid': 'vid_001',
            'clips': [
                {
                    'clip_uid': 'clip_001',
                    'annotations': [
                        {
                            'annotation_uid': 'ann_001',
                            'query_sets': {
                                'qs_001': {},
                                'qs_002': {}
                            }
                        },
                        {
                            'annotation_uid': 'ann_002',
                            'query_sets': {
                                'qs_003': {}
                            }
                        }
                    ]
                }
            ]
        },
        {
            'video_uid': 'vid_002',
            'clips': [
                {
                    'clip_uid': 'clip_002',
                    'annotations': [
                        {
                            'annotation_uid': 'ann_003',
                            'query_sets': {
                                'qs_004': {}
                            }
                        }
                    ]
                }
            ]
        }
    ]
}

# Build predicted_rts dict with some entries
predicted_rts = {
    ('ann_001', 'qs_001'): [PredictedRT()],
    ('ann_001', 'qs_002'): [PredictedRT()],
    ('ann_003', 'qs_004'): [PredictedRT()]
}

ARGS = {
    'annotations': annotations,
    'predicted_rts': predicted_rts
}
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 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
gelu Ran
ran, honoured the contract we drafted
fb-reps/PRVQL/bert_model/bert_module.py
code served (permissive licence) · get_code("69e3a274f98c15af")
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) 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 math
import torch

torch.manual_seed(0)

# The gelu function takes a single parameter 'x' which should be a tensor
# The function applies element-wise operations: multiplication, addition, division, erf
# So x should be a torch tensor of any shape

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
gelu_new Ran
ran, honoured the contract we drafted
fb-reps/PRVQL/bert_model/bert_module.py
pointer only (licence: MIT); the stored copy get_code serves, from lonePatient/BERT-SDA, is pointer only (licence: NONE) · get_code("77601724cad03f95")
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 lonePatient/BERT-SDA, 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.

get_bbox_from_data Ran
ran on a synthesized input
fb-reps/PRVQL/dataset/base_dataset.py
code served (permissive licence) · get_code("070d7eed4f22f51d")
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 int; 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

torch.manual_seed(0)

# The function get_bbox_from_data expects a dict with keys: 'x', 'y', 'width', 'height'
# It accesses these values and performs arithmetic to compute a bounding box

data = {
    'x': 10,
    'y': 20,
    'width': 100,
    'height': 50
}

ARGS = {
    'data': data
}
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 list, ndim 1, dtype family int — 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
get_losses Ran
ran on a synthesized input
fb-reps/PRVQL/utils/loss_utils.py
code served (permissive licence) · get_code("0e96da7c96574812")
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 torch
import torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small config class with the required attributes
class Config:
    def __init__(self):
        self.loss = LossConfig()

class LossConfig:
    def __init__(self):
        self.prob_bce_weight = [1.0, 2.0]
        self.weight_bbox_center = 1.0
        self.weight_bbox_hw = 1.0
        self.weight_bbox_giou = 1.0
        self.weight_prob = 1.0
        self.weight_bbox_ratio = 1.0

# Create config instance
config = Config()

# Create predictions dict with required keys
# Batch size B=2, Time steps T=3, feature dims C=4
B, T, C = 2, 3, 4

preds = {
    'center': torch.randn(B, T, 2),      # (B, T, 2) center coordinates
    'hw': torch.randn(B, T, 2),          # (B, T, 2) height and width
    'bbox': torch.randn(B, T, 4),        # (B, T, 4) bounding boxes (x1, y1, x2, y2)
    'prob': torch.randn(B, T),           # (B, T) probability logits
    'bbox_ratio': torch.randn(B, T),     # (B, T) bbox ratios
}

# Create ground truth dict with required keys
gts = {
    'clip_bbox': torch.rand(B, T, 4) * 100,              # (B, T, 4) bounding boxes
    'clip_with_bbox': torch.randint(0, 2, (B, T)).float(),  # (B, T) binary mask
    'before_query': torch.randint(0, 2, (B, T)).float(),    # (B, T) binary mask
}

# Ensure clip_bbox values are valid (x1 < x2, y1 < y2)
gts['clip_bbox'][:, :, 2:] = gts['clip_bbox'][:, :, :2] + torch.rand(B, T, 2) * 50 + 10

ARGS = {
    'config': config,
    'preds': preds,
    'gts': gts,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, torch — 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
get_losses_head Ran
ran on a synthesized input
fb-reps/PRVQL/utils/loss_utils.py
code served (permissive licence) · get_code("71ace89be308adbc")
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. 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

torch.manual_seed(0)

# Build config object with required attributes
class Config:
    def __init__(self):
        self.model = ModelConfig()
        self.loss = LossConfig()

class ModelConfig:
    def __init__(self):
        self.positive_threshold = 0.5

class LossConfig:
    def __init__(self):
        self.prob_bce_weight = [1.0, 2.0]

config = Config()

# Build batch size b=2, time steps t=3
b, t = 2, 3

# refine_prob shape [b, t]
refine_prob = torch.randn(b, t)

# gts dictionary with required keys
gts = {
    'clip_with_bbox': torch.randint(0, 2, (b, t)).float(),  # shape [b, t]
    'before_query': torch.randint(0, 2, (b, t)).float(),     # shape [b, t]
    'clip_bbox': torch.randn(b, t, 4)                         # shape [b, t, 4]
}

# preds_top dictionary with bbox predictions
preds_top = {
    'bbox': torch.randn(b, t, 4)  # shape [b, t, 4]
}

ARGS = {
    'config': config,
    'refine_prob': refine_prob,
    'gts': gts,
    'preds_top': preds_top
}
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 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: none by the harness; any seed is in the setup block
positionalencoding1d Ran
ran, honoured the contract we drafted
fb-reps/PRVQL/utils/model_utils.py
code served (permissive licence) · get_code("bed1828c9c920325")
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 (4, 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 math

torch.manual_seed(0)

ARGS = {
    "d_model": 8,
    "length": 4
}
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 (4, 8) — shape (4, 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: none by the harness; any seed is in the setup block
positionalencoding3d Ran
ran on a synthesized input
fb-reps/PRVQL/utils/model_utils.py
code served (permissive licence) · get_code("5b821635ccbae36a")
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 (24, 12) 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 numpy as np
from einops import rearrange

torch.manual_seed(0)

ARGS = {
    'd_model': 12,
    'height': 4,
    'width': 3,
    'depth': 2,
    'type': 'sinusoidal'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, numpy, 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 (24, 12) — shape (24, 12) 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
replicate_sample_for_hnm Ran
ran on a synthesized input
fb-reps/PRVQL/dataset/dataset_utils.py
code served (permissive licence) · get_code("379a559fff316790")
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 torch

torch.manual_seed(0)

# Small dimensions to keep tensors tiny
b = 2  # batch size
t = 3  # time steps
c = 4  # channels
h = 5  # height
w = 6  # width

device = 'cpu'

# Construct the gts dictionary with all required keys
gts = {
    'clip': torch.randn(b, t, c, h, w, device=device),
    'clip_origin': torch.randn(b, t, c, h, w, device=device),
    'clip_with_bbox': torch.randint(0, 2, (b, t), dtype=torch.float32, device=device),
    'before_query': torch.randint(0, 2, (b, t), dtype=torch.bool, device=device),
    'clip_bbox': torch.rand(b, t, 4, device=device),
    'query': torch.randn(b, c, h, w, device=device),
    'query_origin': torch.randn(b, c, h, w, device=device),
    'clip_h': torch.tensor([h, h], dtype=torch.long, device=device),
    'clip_w': torch.tensor([w, w], dtype=torch.long, device=device),
}

ARGS = {
    'gts': gts
}
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 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
sample_frames_balance Ran
ran on a synthesized input
fb-reps/PRVQL/dataset/base_dataset.py
code served (permissive licence) · get_code("68900fa3cd5db048")
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 int; 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 random
import numpy as np

# Set seed for reproducibility
random.seed(0)
np.random.seed(0)

ARGS = {
    'num_frames': 5,
    'query_frame': 10,
    'frame_interval': 2,
    'sample': {
        'response_track_valid_range': [5, 25]
    },
    'sampling': 'rand'
}
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 int — 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
positionalencoding2d Not yet run
unverified (harvested, no recorded run)
fb-reps/PRVQL/utils/model_utils.py
code served (permissive licence) · get_code("b2e8731183a22bb6")
How it ran

no invocation recorded.

process_data Not yet run
unverified (harvested, no recorded run)
fb-reps/PRVQL/dataset/dataset_utils.py
code served (permissive licence) · get_code("af0e3990e7f9effe")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 1 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.

FunctionAlso present inEarliest-datedOthers named
gelu 198 other papers
183 dated earlier, 14 later, 1 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
gelu_new 23 other papers
23 dated earlier
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1909.11942 (2019-09)
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
1911.03631 (2019-11)
Hierarchical Graph Network for Multi-hop Question Answering
and 20 more not named here
positionalencoding1d 5 other papers
5 dated earlier
2008.09105 (2020-08)
Location-Aware Graph Convolutional Networks for Video Question Answering
2106.12102 (2021-06)
LegoFormer: Transformers for Block-by-Block Multi-view 3D Reconstruction
2112.06197 (2021-12)
Video as Conditional Graph Hierarchy for Multi-Granular Question Answering
and 2 more not named here

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

Egocentric visual query localization (EgoVQL) focuses on localizing the target of interest in space and time from firstperson videos, given a visual query. Despite recent progressive, existing methods often struggle to handle severe object appearance changes and cluttering background in the video due to lacking sufficient target cues, leading to degradation. Addressing this, we introduce PRVQL, a novel Progressive knowledge-guided Refinement framework for EgoVQL. The core is to continuously exploit target-relevant knowledge directly from videos and utilize it as guidance to refine both query and video features for improving target localization. Our PRVQL contains multiple processing stages. The target knowledge from one stage, comprising appearance and spatial knowledge extracted via two specially designed knowledge learning modules, are utilized as guidance to refine the query and videos features for the next stage, which are used to generate more accurate knowledge for further feature refinement. With such a progressive process, target knowledge in PRVQL can be gradually improved, which, in turn, leads to better refined query and video features for localization in the final stage. Compared to previous methods, our PRVQL, besides the given object cues, enjoys additional crucial target information from a video as guidance to refine features, and hence enhances EgoVQL in complicated scenes. In our experiments on challenging Ego4D, PRVQL achieves stateof-the-art result and largely surpasses other methods, showing its efficacy. Our code, model and results will be released at https://github.com/fb-reps/PRVQL.

For agents

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

get_harvested_code_for_paper("2502.07707")
get_code_for_paper("2502.07707")
have("2502.07707")

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

curl https://syntology.ai/api/ran/2502.07707.json

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

[![Syntology run record](https://syntology.ai/api/ran/2502.07707.svg)](https://syntology.ai/paper/2502.07707)

Connect an agent — have() is free.