The repo supports three additional CBM families beyond the standard CBM/DNN:

• CEM — Concept Embedding Model (--cbm-family cem)
• ProbCBM — Probabilistic Concept Bottleneck Model (--cbm-family probcbm)
• ECBM — Energy-based Concept Bottleneck Model (--cbm-family ecbm)

ECBM is included in the repo and works out of the box (it follows the authors' code). CEM and ProbCBM require the official mateoespinosa/cem package — install it with:

./scripts/install_cem_repo.sh

If you prefer the manual path:

git clone https://github.com/mateoespinosa/cem.git third_party/cem
python -m pip install "pytorch-lightning>=1.6,<2.0" "torchmetrics<1.0"
python -m pip install -r third_party/cem/requirements.txt
python -m pip install -e third_party/cem

Once installed, use --cbm-family to select the model in the robot and sudoku pipelines (the text pipeline trains the CBM only):

# Train and evaluate a CEM on the robot benchmark
python scripts/robot_pipeline.py --seed 1014 --cbm-family cem

# Train and evaluate a ProbCBM on the robot benchmark
python scripts/robot_pipeline.py --seed 1014 --cbm-family probcbm

# Train and evaluate an ECBM on the robot benchmark
python scripts/robot_pipeline.py --seed 1014 --cbm-family ecbm

# Any family on sudoku
python scripts/sudoku_pipeline.py --seed 171 --cbm-family cem
python scripts/sudoku_pipeline.py --seed 171 --cbm-family ecbm

Key configuration options (fields of the config dataclass):

Notes:

• Existing CBM / DNN / conceptual-safeguards paths do not require these packages.
• Every family runs on the robot and sudoku benchmarks.
• Alignment remains on the original cbm path.

This guide shows how to evaluate your own concept bottleneck model on the benchmarks provided by this package. All examples below use the robot benchmark, but the same approach works for sudoku.

Prerequisite: You need the full repository (not just pip install concept-benchmark) to run the pipeline scripts and examples.

Getting data for your model

Generate a dataset as shown in the Quick Start, then access the splits:

train, val, test = dataset.train, dataset.val, dataset.test

Each split is a ConceptDatasetSample with these attributes:

Access formats:

# NumPy arrays; X (also .inputs) holds the image file names, the texts or the feature rows
X_train, C_train, y_train = train.X, train.C, train.y

# PyTorch DataLoader
loader = train.loader(batch_size=64, shuffle=True)
for x_batch, c_batch, y_batch in loader:
    ...

# Pandas DataFrame
df = train.to_dataframe()

Splitting

sample() splits the dataset into train/val/test. Sizes can be absolute counts or fractions:

# Absolute counts
dataset.sample(test_size=10000, val_size=1000, train_size=3800, seed=42)

# Fractions (of total dataset size)
dataset.sample(test_size=0.2, val_size=0.2, seed=42)

# Stratified — preserves class proportions in each split
dataset.sample(test_size=0.2, val_size=0.2, stratify=dataset.y, seed=42)

# Group-based — no group appears in multiple splits (group_ids: one id per row, e.g. the robot's identity)
dataset.sample(test_size=0.2, val_size=0.2, groups=group_ids, seed=42)

# Skewed training set — ensure min-fraction of specific concept patterns
dataset.sample(
    test_size=10000, val_size=0.2, train_size=3800,
    sampling_constraints=[{"concepts": {"my_concept": 1}, "min_fraction": 0.3}],
    seed=42,
)

You can re-split at any time by calling sample() again.

Concept missingness

Simulate missing concept annotations (e.g., incomplete labels from crowdsourcing). Use the generator transforms to produce a new dataset with missingness applied:

from concept_benchmark.transforms import ConceptMissingnessGenerator

dataset.sample(test_size=0.2, val_size=0.2, seed=42)

# MCAR: each concept label independently missing with probability p
dataset = ConceptMissingnessGenerator(dataset, p=0.2, mechanism="mcar", seed=99).generate()

# MNAR: missingness depends on concept value (present concepts more likely observed)
dataset = ConceptMissingnessGenerator(
    dataset, p=0.2, mechanism="mnar", seed=99,
    mnar_config={"present_prob": 0.8, "absent_prob": 0.1},
).generate()

Similarly, ConceptNoiseGenerator adds symmetric or asymmetric label flips, and LabelNoiseGenerator corrupts target labels:

from concept_benchmark.transforms import ConceptNoiseGenerator, LabelNoiseGenerator

dataset = ConceptNoiseGenerator(dataset, p=0.1, seed=99).generate()
dataset = LabelNoiseGenerator(dataset, p=0.05, seed=99).generate()

Wrapping your concept detector

The intervention runner requires a ConceptDetector subclass. Wrap your model:

from experiments.models import ConceptDetector

class MyConceptDetector(ConceptDetector):
    def __init__(self, my_model):
        super().__init__()
        self._my_model = my_model

    def predict_proba(self, dataset, **kwargs):
        """Must return (N, n_concepts) float array in [0, 1]."""
        return self._my_model.predict_concept_probs(dataset.X)

Key point: Override predict_proba() — it receives a ConceptDatasetSample, not raw arrays. Access inputs via dataset.X. The base class predict() calls predict_proba() and thresholds at 0.5 automatically.

Using a PyTorch module directly

If your model is already a PyTorch nn.Module, you can pass it directly instead of subclassing. Your module's forward() must:

• Accept a batched input tensor (e.g. (B, 3, 32, 32) for images)
• Return (B, n_concepts) raw logits (pre-sigmoid) — sigmoid is applied internally by predict()

The return value can also be a tuple (first element is used) or a dict with a "logits" key.

cd = ConceptDetector(model=my_pytorch_module)
cd.fit(train, val, fit_params={"epochs": 50, "lr": 1e-3})

You can also split your model into a backbone + concept head using the embedding_model parameter. The detector will probe the backbone's output shape and attach an MLP head automatically:

cd = ConceptDetector(embedding_model=my_backbone)
cd.fit(train, val, fit_params={"epochs": 50, "lr": 1e-3})

Wrapping your label predictor

Subclass FrontEndModel and override predict() and predict_proba():

from experiments.models import FrontEndModel

class MyFrontEnd(FrontEndModel):
    def __init__(self, my_classifier):
        super().__init__()
        self._clf = my_classifier

    def predict(self, C):
        """C is (N, n_concepts) binary 0/1. Returns (N,) int labels."""
        return self._clf.predict(C)

    def predict_proba(self, C):
        """C is (N, n_concepts) binary 0/1. Returns (N, n_classes) probs."""
        return self._clf.predict_proba(C)

Key point: C is already binarized at 0.5 — binary 0/1 values, not probabilities.

For a simple logistic regression baseline, the built-in FrontEndModel() works out of the box:

fe = FrontEndModel()
fe.fit(train.C, train.y)

Assembling and evaluating

Combine your concept detector and label predictor into a ConceptBasedModel:

import numpy as np
from experiments.models import ConceptBasedModel

cbm = ConceptBasedModel(
    concept_detector=MyConceptDetector(my_concept_model),
    label_predictor=MyFrontEnd(my_classifier),
)

predictions = cbm.predict(test)
accuracy = np.mean(predictions == test.y)
print(f"CBM accuracy: {accuracy:.4f}")

For running interventions and alignment on your model, see the Interventions and Alignment section and examples/robot_pipeline_example.py.

Metrics and Plots

The concept_benchmark.evaluation module provides standalone metric functions and plotting utilities for evaluating CBMs:

from concept_benchmark.evaluation import (
    accuracy, gain, selective_accuracy, coverage,
    plot_intervention_curve, plot_intervention_heatmap, plot_concept_report,
)

Metrics:

Blocks that need a model live in experiments.evaluate (requires cloning the repo); the pipelines are built from them:

Plots take the results tables that the pipelines write (or your own, with the same columns). Rows from several runs are averaged and drawn with a standard-error band, so pass the concatenated results of all your seeds.

Outcome plots show what happens:

Diagnostic plots show why:

import pandas as pd
from concept_benchmark.evaluation import plot_intervention_curve, plot_intervention_heatmap

# one results file per seed, as written by scripts/robot_pipeline.py
results = pd.concat(
    # results_by_seed: {seed: path to a pipeline's results CSV}; dnn_accuracies: one DNN accuracy per seed
    pd.read_csv(path).assign(seed=seed) for seed, path in results_by_seed.items()
)
fig, ax = plot_intervention_curve(results, group="model_family", baseline_accuracy=dnn_accuracies)
fig, ax = plot_intervention_heatmap(results)

All plot functions return (fig, ax). The pipelines' plot stage draws the ones that apply to a run. The examples below use the 10-seed results of the paper (scripts/paper/make_readme_figures.py).

Interventions

Interventions are the core benefit of concept bottleneck models: at test time, a user (or automated system) can inspect and correct the model's concept predictions before the final label is determined.

Oracle interventions (manual approach)

The simplest way to run interventions is to directly manipulate concept predictions. After training a ConceptDetector and FrontEndModel, the workflow is:

1. Get concept probabilities from the detector
2. Identify the most uncertain concepts per sample
3. Replace them with ground-truth values
4. Re-predict with the label model

import numpy as np
from experiments.models import ConceptDetector, FrontEndModel

# cd and fe are a trained concept detector and label predictor, e.g. cbm.concept_detector and
# cbm.label_predictor of a CBM from experiments.evaluate.train_cbm

# Step 1: Get concept probabilities
concept_probs = cd.predict_proba(test)

# Step 2-3: For each sample, replace the k most uncertain concepts
# with ground-truth values
for k in [1, 3]:
    C_intervened = concept_probs.copy()
    uncertainty = np.abs(concept_probs - 0.5)  # distance from decision boundary
    for i in range(len(test)):
        most_uncertain = np.argsort(uncertainty[i])[:k]
        C_intervened[i, most_uncertain] = test.C[i, most_uncertain]

    # Step 4: Threshold to binary and predict
    preds = fe.predict((C_intervened > 0.5).astype(np.float32))
    acc = np.mean(preds == test.y)
    print(f"k={k}: accuracy={acc:.4f}")

Using the intervention API

For more complex intervention scenarios (budgets, strategies, batched evaluation), use the ConceptInterventionRunner:

import numpy as np
from experiments.models import ConceptBasedModel
from experiments.intervention import ConceptInterventionRunner, InterventionConfig
from experiments.kflip import KFlipInterventionStrategy

# Combine detector and label predictor into a CBM
cbm = ConceptBasedModel(concept_detector=cd, label_predictor=fe)

# Configure the intervention
config = InterventionConfig(
    per_instance_budget=3,  # correct up to 3 concepts per sample
    score_threshold=0.2,          # intervene when the label would change
                                  # with probability 0.2 or more
)

# Run interventions
runner = ConceptInterventionRunner(model=cbm)
strategy = KFlipInterventionStrategy()
result = runner.run(strategy, config, test)

# InterventionResult contains y_prob_before/after and y_pred_after
acc_before = np.mean(np.argmax(result.y_prob_before, axis=1) == test.y)
acc_after = np.mean(result.y_pred_after == test.y)
print(f"Accuracy before: {acc_before:.4f}")
print(f"Accuracy after:  {acc_after:.4f}")
print(f"Concepts corrected: {result.mask.sum()}")

Key classes

• InterventionConfig — controls intervention budgets, thresholds, and per-instance caps
• KFlipInterventionStrategy — the default strategy: evaluates all subsets of up to k concepts per sample and selects the intervention that maximizes predicted confidence
• ConceptInterventionRunner — coordinates intervention execution and before/after evaluation

Writing a custom strategy

You can implement your own intervention strategy by subclassing InterventionStrategy and implementing the propose() method.

The intervention flow

When ConceptInterventionRunner.run() is called, it:

1. Builds an InterventionBatch from the dataset (concept predictions + ground truth)
2. Calls strategy.propose(model, batch, config) → returns a StrategyProposal
3. Applies the proposal's mask to replace predicted concepts with ground truth
4. Re-predicts labels with the corrected concepts
5. Returns an InterventionResult with before/after predictions

Key data classes

InterventionBatch — the input your strategy receives:

StrategyProposal — what your strategy returns:

Minimal example

Here's a strategy that intervenes on concepts closest to the 0.5 decision boundary:

import numpy as np
from experiments.intervention import InterventionStrategy, StrategyProposal

class UncertaintyStrategy(InterventionStrategy):
    def __init__(self):
        super().__init__(name="uncertainty")

    def propose(self, model, batch, config):
        k = int(min(config.per_instance_limit(batch.n_concepts), batch.n_concepts))
        mask = np.zeros((batch.n_samples, batch.n_concepts), dtype=bool)

        # Rank concepts by uncertainty (closeness to 0.5)
        uncertainty = 0.5 - np.abs(batch.C_pred - 0.5)
        for i in range(batch.n_samples):
            top_k = np.argsort(uncertainty[i])[-k:]
            mask[i, top_k] = True

        return StrategyProposal(mask=mask)

Use it with the runner:

result = runner.run(
    strategy=UncertaintyStrategy(),
    config=InterventionConfig(per_instance_budget=3, score_threshold=0.2),
    dataset=test,
)

The prepare() hook

Override prepare() if your strategy needs a validation pass before inference — for example, to precompute a global concept ordering:

class MyStrategy(InterventionStrategy):
    def __init__(self):
        super().__init__(name="my_strategy")

    def prepare(self, model, batch, config):
        """Called once on the validation set before run()."""
        # Compute concept importance from validation data
        self._state["concept_order"] = compute_importance(model, batch)

    def propose(self, model, batch, config):
        order = self._state["concept_order"]
        # ... use precomputed order

Call runner.prepare() before runner.run() to trigger the hook:

runner.prepare(strategy, config, validation_dataset=val)
result = runner.run(strategy, config, dataset=test)

Concept sources and intervention sources

The robot pipeline varies two things independently: who annotates the concepts a model is trained on, and who answers when the model asks about a concept at test time.

python scripts/robot_pipeline.py --seed 1014 --concept-preset foot_subtypes \
    --concept-sources human_concepts machine_annotation \
    --intervention-sources perfect expert self --budgets 1 3 max

LLM interventions need an API key (--llm-api-key or GEMINI_API_KEY); --llm-cache-only reuses saved answers.

Regimes

A regime names one pairing of concept source and intervention source, as in the first version of the benchmark:

Both robot pipelines accept --regimes as shorthand for the matching sources:

python scripts/robot_pipeline.py --seed 1014 --concept-preset foot_subtypes --regimes baseline expert subjective
python scripts/robot_text_pipeline.py --seed 1337 --regimes baseline expert subjective

For a complete end-to-end example with training, interventions (through the intervention_table block, which runs ConceptInterventionRunner with the paper's policy) and alignment, see examples/robot_pipeline_example.py.

Alignment

Alignment constraints force the label predictor's concept weights to match a user's prior expectations about concept-label relationships. For example, if a domain expert knows that "has knees" should positively predict the Glorp class, alignment constrains that weight to be positive during retraining.

Why alignment matters

In standard CBM training, the label predictor (logistic regression on concept activations) learns weights freely from data. This can produce counterintuitive weights — e.g., has_knees getting a negative weight even when knees truly indicate Glorp — because the model exploits correlations among imperfect concept predictions.

The paper shows that an alignment constraint can:

• Raise accuracy before interventions — the constrained model no longer relies on a concept its detector reads at chance.
• Remove the benefit of interventions — after every concept is corrected, the unconstrained model is the more accurate one.

The constraint closes the pathway through which a corrected concept reaches the label.

Usage

After training a ConceptBasedModel, use run_alignment() to retrain the frontend with sign constraints and compare:

from experiments.utils import run_alignment

results = run_alignment(
    concept_based_model=cbm,
    train_dataset=train,
    test_dataset=test,
    monotonicity_constraints={"has_knees": 1},  # require a non-negative weight
)

print(f"Original accuracy: {results['original_accuracy']:.4f}")
print(f"Aligned accuracy:  {results['aligned_accuracy']:.4f}")
print(f"Accuracy change:   {results['accuracy_change']:+.4f}")

The monotonicity_constraints dict maps concept names to their required sign: +1 for positive weight, -1 for negative.

Expected results

Mean accuracy over the 10 seeds of the paper (balanced rule):

For a complete end-to-end example with training, interventions, and alignment, see examples/robot_pipeline_example.py.