SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.08469 · 2024

PAL: Pluralistic Alignment Framework for Learning from Heterogeneous Preferences

arXiv · PDF · Open in the Atlas

Code that ran

2 of 4 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; 2 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/2406.08469.json · badge: /api/ran/2406.08469.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
RamyaLab/pluralistic-alignment canonical (classifier label)
classifier basis: census
2 ran · 2 no recorded run
FunctionStatusWhere it lives
preprocess_tokenized_ds Ran
ran; our contract draft was wrong, not the code
RamyaLab/pluralistic-alignment/src/llms_dataset.py
code served (permissive licence) · get_code("ba24ce317191425f")
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 logging

torch.manual_seed(0)

# Set up logging
logging.basicConfig(level=logging.WARNING)

# Create a minimal batch dictionary with tensors
# The function expects input_ids and attention_mask
# batch shape should be (2*bs, seq_len) where bs is batch size
# We'll use bs=2, so shape is (4, 8)

bs = 2
seq_len = 8
PAD_ID = 1

# Create input_ids: shape (4, 8) - 4 sequences of length 8
# First 2 are "chosen", next 2 are "rejected"
input_ids = torch.tensor([
    [0, 2, 3, 4, 1, 1, 1, 1],  # chosen 1
    [0, 5, 6, 7, 1, 1, 1, 1],  # chosen 2
    [0, 2, 8, 9, 1, 1, 1, 1],  # rejected 1 (diverges at index 2)
    [0, 5, 6, 10, 1, 1, 1, 1], # rejected 2 (diverges at index 3)
], dtype=torch.long)

# Create attention_mask: shape (4, 8)
# 1 for real tokens, 0 for padding
attention_mask = torch.tensor([
    [1, 1, 1, 1, 0, 0, 0, 0],
    [1, 1, 1, 1, 0, 0, 0, 0],
    [1, 1, 1, 1, 0, 0, 0, 0],
    [1, 1, 1, 1, 0, 0, 0, 0],
], dtype=torch.long)

batch = {
    'input_ids': input_ids,
    'attention_mask': attention_mask,
}

device = 'cpu'
num_padding_at_beginning = 1

ARGS = {
    'batch': batch,
    'device': device,
    'num_padding_at_beginning': num_padding_at_beginning,
    'seq_len': seq_len,
    'PAD_ID': PAD_ID,
}
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
preprocess_tokenized_ds_rm_pal_b Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
RamyaLab/pluralistic-alignment/src/llms_dataset.py
code served (permissive licence) · get_code("76138c7ab3fe3fa9")
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)

# Create small tensors for the batch
# batch structure: [prompt_ids, prompt_mask, chosen_ids, chosen_mask, rejected_ids, rejected_mask]
# Each has shape (bs, 1, seq_len) based on how they're indexed in the function

bs = 2
seq_len = 512
num_features = 4

# Create token ID tensors with padding
prompt_token_ids = torch.randint(2, 100, (bs, 1, seq_len))
prompt_attention_mask = torch.ones((bs, 1, seq_len))

chosen_token_ids = torch.randint(2, 100, (bs, 1, seq_len))
chosen_attention_mask = torch.ones((bs, 1, seq_len))

rejected_token_ids = torch.randint(2, 100, (bs, 1, seq_len))
rejected_attention_mask = torch.ones((bs, 1, seq_len))

# Add some padding (PAD_ID=1) at the beginning to make padding detection work
for i in range(bs):
    prompt_token_ids[i, 0, :5] = 1
    chosen_token_ids[i, 0, :5] = 1
    rejected_token_ids[i, 0, :5] = 1

batch = [
    prompt_token_ids,
    prompt_attention_mask,
    chosen_token_ids,
    chosen_attention_mask,
    rejected_token_ids,
    rejected_attention_mask
]

ARGS = {
    'batch': batch,
    'device': 'cpu',
    'num_padding_at_beginning': 1,
    'seq_len': seq_len,
    'PAD_ID': 1
}
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
dataset_factory Not yet run
unverified (harvested, no recorded run)
RamyaLab/pluralistic-alignment/src/dataset_factory.py
code served (permissive licence) · get_code("a09eb631c89a5706")
How it ran

no invocation recorded.

get_tokenized_ds_rm_pal_b Not yet run
unverified (harvested, no recorded run)
RamyaLab/pluralistic-alignment/src/llms_dataset.py
code served (permissive licence) · get_code("2232b1c4b70e8413")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 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 4 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

Large foundation models pretrained on raw web-scale data are not readily deployable without additional step of extensive alignment to human preferences. Such alignment is typically done by collecting large amounts of pairwise comparisons from humans ("Do you prefer output A or B?") and learning a reward model or a policy with the Bradley-Terry-Luce (BTL) model as a proxy for a human's underlying implicit preferences. These methods generally suffer from assuming a universal preference shared by all humans, which lacks the flexibility of adapting to plurality of opinions and preferences. In this work, we propose PAL, a framework to model human preference complementary to existing pretraining strategies, which incorporates plurality from the ground up. We propose using the ideal point model as a lens to view alignment using preference comparisons. Together with our novel reformulation and using mixture modeling, our framework captures the plurality of population preferences while simultaneously learning a common preference latent space across different preferences, which can few-shot generalize to new, unseen users. Our approach enables us to use the penultimate-layer representation of large foundation models and simple MLP layers to learn reward functions that are on-par with the existing large state-of-the-art reward models, thereby enhancing efficiency of reward modeling significantly. We show that PAL achieves competitive reward model accuracy compared to strong baselines on 1) Language models with Summary dataset ; 2) Image Generative models with Pick-a-Pic dataset ; 3) A new semisynthetic heterogeneous dataset generated using Anthropic Personas. Finally, our experiments also highlight the shortcoming of current preference datasets that are created using rigid rubrics which wash away heterogeneity, and call for more nuanced data collection approaches.

For agents

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

get_harvested_code_for_paper("2406.08469")
get_code_for_paper("2406.08469")
have("2406.08469")

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

curl https://syntology.ai/api/ran/2406.08469.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.08469.svg)](https://syntology.ai/paper/2406.08469)

Connect an agent — have() is free.