SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2307.02469 · 2023

What Matters in Training a GPT4-Style Language Model with Multimodal Inputs?

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 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, 1 violated it and 2 have no contract verdict; 1 failure 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/2307.02469.json · badge: /api/ran/2307.02469.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
bytedance/lynx-llm canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
fudandisc/reform-eval community (classifier label)
classifier basis not recorded on the link
2 ran · 2 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
disabled_train Ran
ran, violated the contract we drafted
this paper's copy was not recorded; identical code first harvested from zhouc20/LatentGraphDiffusion
pointer only · get_code("4cb732f513d69dfd")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

evaluation Ran
ran; our contract draft was wrong, not the code
bytedance/lynx-llm/generate.py
code served (permissive licence) · get_code("cd56d16a521c2a10")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def generate(self, vision_input, input_ids, input_atts, use_nucleus_sampling=False, 
                 apply_lemmatizer=False, num_beams=1, min_length=1, length_penalty=1.0,
                 no_repeat_ngram_size=-1, top_p=0.9, top_k=3, max_new_tokens=64):
        batch_size = vision_input.shape[0]
        outputs = ["generated text output"] * batch_size
        return outputs
    
    def eval(self):
        return super().eval()

model = MockModel()

data_loader = [
    (
        torch.tensor([0, 1]),
        torch.randn(2, 4),
        torch.randint(0, 100, (2, 5)),
        torch.ones(2, 5)
    ),
    (
        torch.tensor([2, 3]),
        torch.randn(2, 4),
        torch.randint(0, 100, (2, 5)),
        torch.ones(2, 5)
    )
]

device = 'cpu'

config = {
    'apply_lemmatizer': False,
    'num_beams': 3,
    'min_length': 5,
    'use_nucleus_sampling': False,
    'length_penalty': 1.0,
    'no_repeat_ngram_size': 2,
    'top_p': 0.9,
    'top_k': 5,
    'max_new_tokens': 64
}

ARGS = {
    'model': model,
    'data_loader': data_loader,
    'device': device,
    'config': config
}
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 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
half_new_maybe_autocast Ran
ran on a synthesized input
fudandisc/reform-eval/models/interfaces/blip2.py
code served (permissive licence) · get_code("1421a402c2b95915")
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 nullcontext. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 contextlib

torch.manual_seed(0)

# Create a minimal self object with a device attribute
class MinimalModule:
    def __init__(self, device):
        self.device = device

# Test case 1: CPU device
self_cpu = MinimalModule(torch.device('cpu'))

# Test case 2: CUDA device (will use CPU since CUDA not available, but device attribute is set)
self_cuda = MinimalModule(torch.device('cuda' if torch.cuda.is_available() else 'cpu'))

# Use CPU device for the actual test
self = MinimalModule(torch.device('cpu'))

ARGS = {
    'self': self,
    'dtype': torch.float16
}
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 nullcontext — a nullcontext output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a nullcontext 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
new_maybe_autocast Ran
ran on a synthesized input
fudandisc/reform-eval/models/interfaces/blip2.py
code served (permissive licence) · get_code("3fda0878c53efad7")
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 nullcontext. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 contextlib

torch.manual_seed(0)

class MockModel(torch.nn.Module):
    def __init__(self):
        super().__init__()
        self.device = torch.device('cpu')
    
    def forward(self, x):
        return x

model = MockModel()

ARGS = {
    'self': model,
    'dtype': torch.float32
}
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 nullcontext — a nullcontext output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a nullcontext 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_all_output_names Not yet run
unverified (harvested, no recorded run)
fudandisc/reform-eval/run_eval.py
code served (permissive licence) · get_code("330f46157d56cbf4")
How it ran

no invocation recorded.

get_output_name Not yet run
unverified (harvested, no recorded run)
fudandisc/reform-eval/run_eval.py
code served (permissive licence) · get_code("5ebd28fe8762fc38")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 1 later) and 5 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
disabled_train 173 other papers
30 dated earlier, 5 the same month, 138 later · 5 only as an assembly brick
2012.09841 (2020-12)
Taming Transformers for High-Resolution Image Synthesis
2102.05379 (2021-02)
Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
and 170 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

Recent advancements in Large Language Models (LLMs) such as GPT4 have displayed exceptional multi-modal capabilities in following open-ended instructions given images. However, the performance of these models heavily relies on design choices such as network structures, training data, and training strategies, and these choices have not been extensively discussed in the literature, making it difficult to quantify progress in this field. To address this issue, this paper presents a systematic and comprehensive study, quantitatively and qualitatively, on training such models. We implement over 20 variants with controlled settings. Concretely, for network structures, we compare different LLM backbones and model designs. For training data, we investigate the impact of data and sampling strategies. For instructions, we explore the influence of diversified prompts on the instruction-following ability of the trained models. For benchmarks, we contribute the first, to our best knowledge, comprehensive evaluation set including both image and video tasks through crowd-sourcing. Based on our findings, we present Lynx, which performs the most accurate multi-modal understanding while keeping the best multi-modal generation ability compared to existing open-sourced GPT4-style models.

For agents

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

get_harvested_code_for_paper("2307.02469")
get_code_for_paper("2307.02469")
have("2307.02469")

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

curl https://syntology.ai/api/ran/2307.02469.json

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

[![Syntology run record](https://syntology.ai/api/ran/2307.02469.svg)](https://syntology.ai/paper/2307.02469)

Connect an agent — have() is free.