SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.04809 · ICLR · 2023

Learning Human-Compatible Representations for Case-Based Decision Support

Yuxin Chen, Shi Feng, Han Liu, Yizhou Tian, Chacha Chen, Chenhao Tan

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 (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 3 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/2303.04809.json · badge: /api/ran/2303.04809.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
chicagohai/learning-human-compatible-representations canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
load_configs Ran
ran; our contract draft was wrong, not the code
chicagohai/learning-human-compatible-representations/src/trainer.py
code served (permissive licence) · get_code("ef755af844198952")
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 DictConfig. 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 tempfile
import os
from omegaconf import OmegaConf as oc

# Create temporary config files
temp_dir = tempfile.mkdtemp()

# Base config
base_config_path = os.path.join(temp_dir, 'base_config.yaml')
with open(base_config_path, 'w') as f:
    f.write("""
wandb_group: test_group
wandb_mode: disabled
wandb_project: test_project
wandb_entity: test_entity
wandb_name: test_run
gpus: 1
seed: 42
deterministic: true
dataloader_num_workers: 0
do_train: true
do_test: true
checkpoint_callback: true
enable_progress_bar: false
log_every_n_steps: 10
max_epochs: 5
learning_rate: 0.001
check_val_every_n_epoch: 1
train_batch_size: 32
triplet_batch_size: 16
train_dir: /tmp/train
valid_dir: /tmp/valid
test_dir: /tmp/test
transform: null
aug: null
num_class: 10
syn: false
train_triplets: /tmp/train_triplets.txt
valid_triplets: /tmp/valid_triplets.txt
test_triplets: /tmp/test_triplets.txt
filtered: true
model: resnet50
embed_dim: 128
pretrained: true
lamda: 0.5
embeds_output_dir: /tmp/embeds
test_ckpt_path: /tmp/checkpoint.pt
out_csv: /tmp/output.csv
predicted_labels: /tmp/labels.txt
""")

# Model config
model_config_path = os.path.join(temp_dir, 'model_config.yaml')
with open(model_config_path, 'w') as f:
    f.write("""
model: resnet50
embed_dim: 128
pretrained: true
lamda: 0.5
""")

# Dataset config (optional)
dataset_config_path = os.path.join(temp_dir, 'dataset_config.yaml')
with open(dataset_config_path, 'w') as f:
    f.write("""
train_dir: /tmp/train
valid_dir: /tmp/valid
test_dir: /tmp/test
transform: null
aug: null
num_class: 10
syn: false
""")

# Triplet config (optional)
triplet_config_path = os.path.join(temp_dir, 'triplet_config.yaml')
with open(triplet_config_path, 'w') as f:
    f.write("""
train_triplets: /tmp/train_triplets.txt
valid_triplets: /tmp/valid_triplets.txt
test_triplets: /tmp/test_triplets.txt
filtered: true
""")

# Overwrite config (optional)
overwrite_config_path = os.path.join(temp_dir, 'overwrite_config.yaml')
with open(overwrite_config_path, 'w') as f:
    f.write("""
""")

# Create a simple args object
class Args:
    def __init__(self):
        self.base_config = base_config_path
        self.model_config = model_config_path
        self.dataset_config = dataset_config_path
        self.triplet_config = triplet_config_path
        self.overwrite_config = overwrite_config_path

ARGS = {
    'args': Args()
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, omegaconf — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DictConfig — a DictConfig output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DictConfig output the recorded ndim 1 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
load_configs_sweep Ran
ran; our contract draft was wrong, not the code
chicagohai/learning-human-compatible-representations/src/trainer.py
code served (permissive licence) · get_code("0c9ef28bc83ecaa7")
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 DictConfig. 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 tempfile
import os
from omegaconf import OmegaConf as oc

# Create temporary directory for config files
tmpdir = tempfile.mkdtemp()

# Create base config
base_config_dict = {
    'wandb_group': 'test_group',
    'wandb_mode': 'disabled',
    'wandb_project': 'test_project',
    'wandb_entity': 'test_entity',
    'wandb_name': 'test_run',
    'gpus': 1,
    'seed': 42,
    'deterministic': True,
    'dataloader_num_workers': 0,
    'do_train': True,
    'do_test': True,
    'checkpoint_callback': True,
    'enable_progress_bar': False,
    'log_every_n_steps': 10,
    'max_epochs': 5,
    'learning_rate': 0.001,
    'check_val_every_n_epoch': 1,
    'train_batch_size': 32,
    'triplet_batch_size': 16,
    'train_dir': '/tmp/train',
    'valid_dir': '/tmp/valid',
    'test_dir': '/tmp/test',
    'transform': 'default',
    'aug': True,
    'num_class': 10,
    'syn': False,
    'train_triplets': '/tmp/train_triplets.txt',
    'valid_triplets': '/tmp/valid_triplets.txt',
    'test_triplets': '/tmp/test_triplets.txt',
    'filtered': True,
    'model': 'resnet50',
    'embed_dim': 128,
    'pretrained': True,
    'lamda': 0.5,
    'embeds_output_dir': '/tmp/embeds',
    'test_ckpt_path': '/tmp/checkpoint.pt',
    'out_csv': '/tmp/output.csv',
    'predicted_labels': '/tmp/labels.txt'
}

base_config_path = os.path.join(tmpdir, 'base_config.yaml')
oc.save(base_config_dict, base_config_path)

# Create model config
model_config_dict = {
    'model': 'resnet50',
    'embed_dim': 128,
    'pretrained': True,
    'lamda': 0.5
}

model_config_path = os.path.join(tmpdir, 'model_config.yaml')
oc.save(model_config_dict, model_config_path)

# Create dataset config
dataset_config_dict = {
    'train_dir': '/tmp/train',
    'valid_dir': '/tmp/valid',
    'test_dir': '/tmp/test',
    'transform': 'default',
    'aug': True,
    'num_class': 10,
    'syn': False
}

dataset_config_path = os.path.join(tmpdir, 'dataset_config.yaml')
oc.save(dataset_config_dict, dataset_config_path)

# Create triplet config
triplet_config_dict = {
    'train_triplets': '/tmp/train_triplets.txt',
    'valid_triplets': '/tmp/valid_triplets.txt',
    'test_triplets': '/tmp/test_triplets.txt',
    'filtered': True
}

triplet_config_path = os.path.join(tmpdir, 'triplet_config.yaml')
oc.save(triplet_config_dict, triplet_config_path)

# Create args object
class Args:
    def __init__(self):
        self.base_config = base_config_path
        self.model_config = model_config_path
        self.dataset_config = dataset_config_path
        self.triplet_config = triplet_config_path
        self.learning_rate = -1
        self.train_batch_size = -1
        self.seed = 123

ARGS = {
    'args': Args()
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, omegaconf — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DictConfig — a DictConfig output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DictConfig output the recorded ndim 1 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
test_configs Ran
ran; our contract draft was wrong, not the code
chicagohai/learning-human-compatible-representations/src/trainer.py
code served (permissive licence) · get_code("ec6acd430c3b43b4")
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 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

ARGS = {
    'configs': {
        # wandb_args
        'wandb_group': 'test_group',
        'wandb_mode': 'offline',
        'wandb_project': 'test_project',
        'wandb_entity': 'test_entity',
        'wandb_name': 'test_run',
        
        # trainer_args
        'gpus': 1,
        'seed': 42,
        'deterministic': True,
        'dataloader_num_workers': 0,
        'do_train': True,
        'do_test': True,
        'checkpoint_callback': True,
        'enable_progress_bar': False,
        'log_every_n_steps': 10,
        
        # training_args
        'max_epochs': 5,
        'learning_rate': 0.001,
        'check_val_every_n_epoch': 1,
        'train_batch_size': 32,
        'triplet_batch_size': 16,
        
        # dataset_args
        'train_dir': '/tmp/train',
        'valid_dir': '/tmp/valid',
        'test_dir': '/tmp/test',
        'transform': None,
        'aug': False,
        'num_class': 10,
        'syn': False,
        
        # triplet_args
        'train_triplets': None,
        'valid_triplets': None,
        'test_triplets': None,
        'filtered': False,
        
        # model_args
        'model': 'resnet50',
        'embed_dim': 128,
        'pretrained': True,
        'lamda': 0.5,
        
        # file_args
        'embeds_output_dir': '/tmp/embeds',
        'test_ckpt_path': '/tmp/checkpoint.pt',
        'out_csv': '/tmp/output.csv',
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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

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

Algorithmic case-based decision support provides examples to aid people in decision making tasks by providing contexts for a test case. Despite the promising performance of supervised learning, representations learned by supervised models may not align well with human intuitions: what models consider similar examples can be perceived as distinct by humans. As a result, they have limited effectiveness in case-based decision support. In this work, we incorporate ideas from metric learning with supervised learning to examine the importance of alignment for effective decision support. In addition to instance-level labels, we use human-provided triplet judgments to learn human-compatible decision-focused representations. Using both synthetic data and human subject experiments in multiple classification tasks, we demonstrate that such representation is better aligned with human perception than representation solely optimized for classification. Human-compatible representations identify nearest neighbors that are perceived as more similar by humans and allow humans to make more accurate predictions, leading to substantial improvements in human decision accuracies (17.8% in butterfly vs. moth classification and 13.2% in pneumonia classification).

For agents

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

get_harvested_code_for_paper("2303.04809")
get_code_for_paper("2303.04809")
have("2303.04809")

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

curl https://syntology.ai/api/ran/2303.04809.json

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

[![Syntology run record](https://syntology.ai/api/ran/2303.04809.svg)](https://syntology.ai/paper/2303.04809)

Connect an agent — have() is free.