modaic/hallucination
5
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

(no commit message)

Browse Source
This commit is contained in:
Tyrin Todd
2026-02-10 01:32:45 -08:00
parent 670bbe0db6
commit 85ef8d8b35
3 changed files with 415 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

312
README.md
View File

@@ -1,2 +1,312 @@
# hallucination
# Bench
Modaic internal SDK for benchmarking judges and training confidence probes.
## Installation
```bash
cd cli
uv sync
```
## CLI Commands
All commands are run from the `cli` directory via `uv run mo <command>`.
### `create`
Create benchmark datasets for training confidence probes. This command runs a judge on examples, extracts embeddings via Modal, and pushes the resulting dataset to HuggingFace Hub.
**Subcommands:**
- `create ppe` - Create dataset from PPE (human-preference + correctness) benchmarks
- `create judge_bench` - Create dataset from the JudgeBench benchmark
**Usage:**
```bash
# Interactive mode (recommended) - prompts for configuration
uv run mo create ppe
uv run mo create judge_bench
# With config file
uv run mo create ppe --config config.yaml
uv run mo create judge_bench --config config.yaml
```
**Options:**
| Option | Short | Description |
| ---------- | ----- | -------------------------- |
| `--config` | `-c` | Path to config file (YAML) |
**Config File Example:**
```yaml
judge: tyrin/ppe-judge-gepa
output: tytodd/my-probe-dataset
n_train: 500
n_test: 100
embedding_layer: -1 # -1 for middle layer
```
**What it does:**
1. Loads examples from the benchmark dataset
2. Runs the specified judge on each example to get predictions
3. Extracts embeddings from the judge's LLM via Modal (GPU)
4. Creates a HuggingFace dataset with columns: `question`, `response_a`, `response_b`, `label`, `predicted`, `messages`, `embeddings`
5. Pushes to HuggingFace Hub
---
### `train`
Train a confidence probe on an embeddings dataset created with `create`.
**Usage:**
```bash
# Interactive mode (recommended) - prompts for all configuration
uv run mo train
# With config file
uv run mo train --config config.yaml
# With CLI arguments
uv run mo train --dataset tytodd/my-embeddings --epochs 10 --lr 0.0001
```
**Options:**
| Option | Short | Description | Default |
| ---------------- | ----- | --------------------------------------------------------------------------------- | ----------------- |
| `--config` | `-c` | Path to config file (YAML) | - |
| `--dataset` | `-d` | Dataset path (HuggingFace Hub or local) (must be a dataset created with `create`) | - |
| `--model-path` | `-m` | Output path for trained model | `{dataset}_probe` |
| `--batch-size` | | Batch size | 4 |
| `--epochs` | | Number of training epochs | 10 |
| `--lr` | | Learning rate | 0.0001 |
| `--weight-decay` | | Weight decay | 0.01 |
| `--test-size` | | Validation split ratio (if no test split) | 0.2 |
| `--seed` | | Random seed | 42 |
| `--project` | | W&B project name | model_path |
| `--hub-path` | | HuggingFace Hub path to push model | - |
**Config File Example:**
```yaml
dataset_path: tytodd/my-probe-dataset
model_path: ./best_probe
hub_path: tytodd/my-probe # Optional: push to HF Hub
batch_size: 4
epochs: 10
learning_rate: 0.0001
weight_decay: 0.01
test_size: 0.2
seed: 42
```
**What it does:**
1. Loads an embeddings dataset (from HuggingFace Hub or local)
2. Creates binary labels: 1 if `predicted == label`, 0 otherwise
3. Trains a linear probe using MSE loss (Brier score optimization)
4. Logs metrics to Weights & Biases (Brier, ECE, MCE, Kuiper, AUROC)
5. Saves the best model based on validation Brier score
6. Optionally pushes to HuggingFace Hub
---
### `eval`
Evaluate a trained confidence probe on a dataset. Computes calibration and discrimination metrics.
**Usage:**
```bash
# Interactive mode (recommended) - prompts for probe and dataset
uv run mo eval
# With CLI arguments
uv run mo eval --probe tytodd/my-probe --dataset tytodd/my-embeddings
# Evaluate on train split instead of test
uv run mo eval --probe tytodd/my-probe --dataset tytodd/my-embeddings --split train
```
**Options:**
| Option | Short | Description | Default |
| ---------------------------- | ----- | ---------------------------------------- | ------------ |
| `--probe` | `-p` | Probe path (HuggingFace Hub or local) | - |
| `--dataset` | `-d` | Dataset path (HuggingFace Hub or local) | - |
| `--split` | `-s` | Dataset split to evaluate on | test |
| `--batch-size` | `-b` | Batch size for evaluation | 64 |
| `--normalize/--no-normalize` | `-n` | Normalize embeddings with StandardScaler | probe config |
**Metrics computed:**
| Metric | Description |
| ----------- | ------------------------------------------------- |
| Brier Score | Mean squared error between predictions and labels |
| Accuracy | Classification accuracy at 0.5 threshold |
| F1 Score | Harmonic mean of precision and recall |
| ECE | Expected Calibration Error (10 bins) |
| MCE | Maximum Calibration Error |
| Kuiper | Kuiper statistic for calibration |
| AUROC | Area Under the ROC Curve (discrimination) |
**What it does:**
1. Loads a pretrained probe from HuggingFace Hub or local path
2. Loads a dataset created with `create`
3. Creates binary labels: 1 if `predicted == label`, 0 otherwise
4. Runs inference and computes calibration/discrimination metrics
5. Displays results in a formatted table
---
### `compile`
Compile (optimize) a judge using GEPA over a dataset. GEPA iteratively improves the judge's prompt based on training examples.
**Subcommands:**
- `compile` (base) - Compile with custom dataset and parameter mapping
- `compile ppe` - Compile specifically for PPE datasets (human-preference + correctness)
**Usage:**
```bash
# Interactive mode
uv run mo compile
uv run mo compile ppe
# With config file
uv run mo compile --config config.yaml
uv run mo compile ppe --config config.yaml
```
**Options:**
| Option | Short | Description |
| ---------- | ----- | -------------------------- |
| `--config` | `-c` | Path to config file (YAML) |
**Config File Example:**
```yaml
judge: tyrin/ppe-judge
dataset: tytodd/ppe-human-preference
inputs: # selects which input columns of the dataset to use (not necearry if using a compile subcommand like ppe or judge_bench)
- name: question
- name: response_a
column: response_A # Map param name to dataset column
- name: response_b
column: response_B
label_column: label
n_train: 100
n_val: 50
base_model: gpt-4o-mini
reflection_model: gpt-4o
output: tyrin/ppe-judge-gepa
seed: 42
```
**What it does:**
1. Loads a judge from Modaic Hub
2. Loads training/validation examples from a HuggingFace dataset
3. Maps judge parameters to dataset columns
4. Runs GEPA optimization to improve the judge's prompt
5. Pushes the optimized judge to Modaic Hub
---
### `embed`
Regenerate embeddings for an existing dataset using a different model or layer. Useful for experimenting with different embedding configurations without re-running the judge.
**Usage:**
```bash
# Interactive mode
uv run mo embed
# With CLI arguments
uv run mo embed --dataset tytodd/my-dataset --hf-model Qwen/Qwen3-VL-32B-Instruct --layer -1
```
**Options:**
| Option | Short | Description |
| ------------ | ----- | ---------------------------------------- |
| `--dataset` | `-d` | Dataset path (HuggingFace Hub or local) |
| `--hf-model` | `-m` | HuggingFace model path for embeddings |
| `--layer` | `-l` | Hidden layer index (-1 for middle layer) |
**What it does:**
1. Loads an existing dataset (must have a `messages` column)
2. Regenerates embeddings using the specified model/layer via Modal
3. Replaces the `embeddings` column in the dataset
4. Prompts to push the updated dataset to HuggingFace Hub
**Example workflow:**
```bash
# Original dataset was created with layer 32
# Now try middle layer instead
uv run mo embed \
--dataset tytodd/my-embeddings \
--hf-model Qwen/Qwen3-VL-32B-Instruct \
--layer -1
```
---
## Recommended Embedding Layers
When extracting embeddings, use these recommended layer indices for best probe performance:
| Model | HuggingFace Path | Recommended Layer |
| ------------- | ----------------------------------- | ----------------- |
| GPT-OSS 20B | `openai/gpt-oss-20b` | 8 |
| Qwen3-VL 32B | `Qwen/Qwen3-VL-32B-Instruct` | 16 |
| Llama 3.3 70B | `meta-llama/Llama-3.3-70B-Instruct` | 32 |
Use `-1` for the middle layer if experimenting with an unlisted model.
---
## Typical Workflow
```bash
# 1. Create a probe dataset from a benchmark
uv run mo create ppe
# 2. Train a confidence probe
uv run mo train --dataset tytodd/ppe-qwen3-embeddings
# 3. Evaluate the probe on a test set
uv run mo eval --probe tytodd/my-probe --dataset tytodd/ppe-qwen3-embeddings
# 4. (Optional) Compile/optimize a judge with GEPA
uv run mo compile ppe
# 5. (Optional) Re-embed with different layer
uv run mo embed --dataset tytodd/my-dataset --layer 32
```
## Environment Variables
Create a `.env` file with:
```bash
OPENAI_API_KEY=...
WANDB_API_KEY=...
HF_TOKEN=...
MODAIC_TOKEN=...
TOGETHER_API_KEY=...
```

56
config.json Normal file
View File

@@ -0,0 +1,56 @@
{
"model": null,
"signature": {
"description": "Given ground truth knowledge, a question/context, and a response, determine if the response\ncontains hallucinations (information not supported by the knowledge).\n\nTask: Check if the response accurately reflects the provided knowledge or if it\ncontains fabricated/incorrect information.\n\n- factual: The response is accurate according to the provided knowledge\n- hallucinated: The response contains information not supported by or contradicting the knowledge\n\nConsider:\n1. Does the response make claims not found in the knowledge?\n2. Does the response contradict any facts in the knowledge?\n3. Does the response fabricate details (names, dates, relationships)?\n\nFirst reason through your thought process in the `reasoning` field.\nBe sure to verbalize any uncertainty in your thought process.\nThen output your conclusion in the `label` field.",
"properties": {
"knowledge": {
"__dspy_field_type": "input",
"desc": "Ground truth knowledge/facts to check against",
"prefix": "Knowledge:",
"title": "Knowledge",
"type": "string"
},
"question": {
"__dspy_field_type": "input",
"default": "",
"desc": "The question or context",
"prefix": "Question:",
"title": "Question",
"type": "string"
},
"response": {
"__dspy_field_type": "input",
"desc": "The response to evaluate for hallucination",
"prefix": "Response:",
"title": "Response",
"type": "string"
},
"reasoning": {
"__dspy_field_type": "output",
"desc": "Your step by step reasoning comparing response to knowledge. Verbalize uncertainty.",
"prefix": "Reasoning:",
"title": "Reasoning",
"type": "string"
},
"label": {
"__dspy_field_type": "output",
"desc": "Whether the response is 'factual' or 'hallucinated'",
"enum": [
"factual",
"hallucinated"
],
"prefix": "Label:",
"title": "Label",
"type": "string"
}
},
"required": [
"knowledge",
"response",
"reasoning",
"label"
],
"title": "HallucinationSig",
"type": "object"
}
}

48
program.json Normal file
View File

@@ -0,0 +1,48 @@
{
"traces": [],
"train": [],
"demos": [],
"signature": {
"instructions": "Given ground truth knowledge, a question/context, and a response, determine if the response\ncontains hallucinations (information not supported by the knowledge).\n\nTask: Check if the response accurately reflects the provided knowledge or if it\ncontains fabricated/incorrect information.\n\n- factual: The response is accurate according to the provided knowledge\n- hallucinated: The response contains information not supported by or contradicting the knowledge\n\nConsider:\n1. Does the response make claims not found in the knowledge?\n2. Does the response contradict any facts in the knowledge?\n3. Does the response fabricate details (names, dates, relationships)?\n\nFirst reason through your thought process in the `reasoning` field.\nBe sure to verbalize any uncertainty in your thought process.\nThen output your conclusion in the `label` field.",
"fields": [
{
"prefix": "Knowledge:",
"description": "Ground truth knowledge/facts to check against"
},
{
"prefix": "Question:",
"description": "The question or context"
},
{
"prefix": "Response:",
"description": "The response to evaluate for hallucination"
},
{
"prefix": "Reasoning:",
"description": "Your step by step reasoning comparing response to knowledge. Verbalize uncertainty."
},
{
"prefix": "Label:",
"description": "Whether the response is 'factual' or 'hallucinated'"
}
]
},
"lm": {
"model": "together_ai/Qwen/Qwen3-VL-32B-Instruct",
"model_type": "chat",
"cache": true,
"num_retries": 3,
"finetuning_model": null,
"launch_kwargs": {},
"train_kwargs": {},
"temperature": null,
"max_tokens": null
},
"metadata": {
"dependency_versions": {
"python": "3.11",
"dspy": "3.1.2",
"cloudpickle": "3.1"
}
}
}