SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.04286 · 2024

Bias Amplification in Language Model Evolution: An Iterated Learning Perspective

arXiv · PDF · Open in the Atlas

Code that ran

10 of 10 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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 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/2404.04286.json · badge: /api/ran/2404.04286.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
joshua-ren/iicl canonical (classifier label)
classifier basis: census
10 ran · 0 no recorded run
FunctionStatusWhere it lives
cnt_of_status Ran
ran, honoured the contract we drafted
joshua-ren/iicl/utils/h_and_d.py
code served (permissive licence) · get_code("a675bd17e9f3f27f")
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, ndim 1, dtype family int; 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 numpy as np

torch.manual_seed(0)

# The function cnt_of_status takes a stat_list parameter
# and counts occurrences of 'on', 'off', and other values
# It iterates through stat_list and compares each element to strings

# Create a valid stat_list with mixed status values
stat_list = ['on', 'off', 'on', 'unknown', 'on', 'off', 'off', 'pending']

ARGS = {
    'stat_list': stat_list
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple 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
convert_chatcompl_to_json Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/evaluations.py
code served (permissive licence) · get_code("01b21d0a9ec85282")
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 list, ndim 2, 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

class TopLogprob:
    def __init__(self, token, logprob, bytes_val):
        self.token = token
        self.logprob = logprob
        self.bytes = bytes_val

class ChatCompletion:
    def __init__(self, token, logprob, top_logprobs_list):
        self.token = token
        self.logprob = logprob
        self.top_logprobs = top_logprobs_list

class ChatCompletionChoice:
    def __init__(self, content):
        self.content = content

torch.manual_seed(0)

# Build top_logprobs for first completion
top_logprobs_1 = [
    TopLogprob(token="a", logprob=-0.5, bytes_val=[97]),
    TopLogprob(token="b", logprob=-0.7, bytes_val=[98])
]

# Build completions for first choice
completions_1 = [
    ChatCompletion(token="hello", logprob=-0.1, top_logprobs_list=top_logprobs_1),
    ChatCompletion(token="world", logprob=-0.2, top_logprobs_list=top_logprobs_1)
]

# Build choice with content
choice_1 = ChatCompletionChoice(content=completions_1)

# Build top_logprobs for second completion
top_logprobs_2 = [
    TopLogprob(token="x", logprob=-0.3, bytes_val=[120]),
    TopLogprob(token="y", logprob=-0.4, bytes_val=[121])
]

# Build completions for second choice
completions_2 = [
    ChatCompletion(token="foo", logprob=-0.15, top_logprobs_list=top_logprobs_2),
    ChatCompletion(token="bar", logprob=-0.25, top_logprobs_list=top_logprobs_2)
]

# Build choice with content
choice_2 = ChatCompletionChoice(content=completions_2)

# Construct list_probs argument
list_probs = [choice_1, choice_2]

ARGS = {
    "list_probs": list_probs
}
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 list, ndim 2, 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
dstr_to_pairs Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/evaluations.py
code served (permissive licence) · get_code("1d025eb199f9b305")
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 list. 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.

# Analyze the function: dstr_to_pairs parses a formatted string into pairs
# The string format is: "input: val1, val2, ...\noutput: val\n..."
# The function expects a newline-separated string with alternating input/output lines

d_str = """input: apple, banana, cherry
output: fruit
input: car, truck, bus
output: vehicle"""

ARGS = {
    "d_str": d_str
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list 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
eval_feedback_h Ran
ran, honoured the contract we drafted
joshua-ren/iicl/utils/evaluations.py
code served (permissive licence) · get_code("ac629ac9b68defd2")
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, ndim 1, dtype family int; 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

torch.manual_seed(0)

# Analyze the function:
# - h: dict that is iterated over with .keys() and indexed with [key]
# - fb_h: dict that is checked with .keys() and indexed with [key]
# - N: int used in comparison (corr_cnt == N)
# The function compares dictionaries of predicted rules

# Create sample rule dictionaries
# Using strings as keys to represent rule identifiers
h = {
    'rule_0': 0.95,
    'rule_1': 0.87,
    'rule_2': 0.76,
}

fb_h = {
    'rule_0': 0.95,
    'rule_1': 0.87,
    'rule_3': 0.92,
}

N = 2

ARGS = {
    'h': h,
    'fb_h': fb_h,
    'N': N,
}
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, ndim 1, dtype family int — a tuple 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
gen_data_prompt Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/standard_prompts.py
code served (permissive licence) · get_code("c2ff65785c5bb719")
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 str. 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 pandas as pd
import torch

torch.manual_seed(0)

# Create a simple DataFrame with the required columns
data = pd.DataFrame({
    'obj_str': ['object1', 'object2'],
    'stat': ['stat_result1', 'stat_result2']
})

ARGS = {
    'd': data,
    'need_stat': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
gen_dh_prompt Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/standard_prompts.py
code served (permissive licence) · get_code("b88b11eb13d05a63")
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 str. 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

torch.manual_seed(0)

ARGS = {
    'M': 5,
    'rule': 'If any object is on, output on. If all are off, output off. Otherwise undetermined.'
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
gen_hd_prompt Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/standard_prompts.py
code served (permissive licence) · get_code("158545d133c04f95")
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 str. 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

# Define ARGS dict with valid arguments for gen_hd_prompt
ARGS = {
    "data": "Input: [1, 2, 3]\nOutput: 6\n\nInput: [2, 3, 4]\nOutput: 9\n",
    "ask_rule": "Rule format: If input is [a, b, c], output is a + b + c"
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
gen_hstar_given_status Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/h_and_d.py
code served (permissive licence) · get_code("17a5c87aa0811333")
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 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 torch
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - status: indexed by integer j in range(len(OBJECTS)), so it needs to be indexable
# - OBJECTS: has len() and is indexable by integer j
# - The function creates a dict mapping OBJECTS[j] to status[j]
# - So status and OBJECTS should have the same length

# Create simple test data
status = [0.1, 0.2, 0.3, 0.4]
OBJECTS = ["obj_a", "obj_b", "obj_c", "obj_d"]

ARGS = {
    "status": status,
    "OBJECTS": OBJECTS
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — 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
gen_hstar_rnd Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
joshua-ren/iicl/utils/h_and_d.py
code served (permissive licence) · get_code("cda033e38ec44fbd")
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 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

STATES = ['on', 'off', 'und']

OBJECTS = ['obj1', 'obj2', 'obj3', 'obj4']
PROBS = [0.4, 0.4, 0.2]

ARGS = {
    'OBJECTS': OBJECTS,
    'PROBS': PROBS
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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
get_fblist_from_hdfeedback Ran
ran; our contract draft was wrong, not the code
joshua-ren/iicl/utils/acronym_utils.py
code served (permissive licence) · get_code("f6926d522e90289e")
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 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

# Construct a sample hd_feedback string that contains multiple lines with JSON-like objects
hd_feedback = """{
  "feedback": "good"
}
some random text
{
  "score": 9}
another line
{
  "data": [1, 2, 3]
}"""

ARGS = {
    "hd_feedback": hd_feedback
}
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 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

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 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 10 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

With the widespread adoption of Large Language Models (LLMs), the prevalence of iterative interactions among these models is anticipated to increase. Notably, recent advancements in multi-round self-improving methods allow LLMs to generate new examples for training subsequent models. At the same time, multi-agent LLM systems, involving automated interactions among agents, are also increasing in prominence. Thus, in both short and long terms, LLMs may actively engage in an evolutionary process. We draw parallels between the behavior of LLMs and the evolution of human culture, as the latter has been extensively studied by cognitive scientists for decades. Our approach involves leveraging Iterated Learning (IL), a Bayesian framework that elucidates how subtle biases are magnified during human cultural evolution, to explain some behaviors of LLMs. This paper outlines key characteristics of agents' behavior in the Bayesian-IL framework, including predictions that are supported by experimental verification with various LLMs. This theoretical framework could help to more effectively predict and guide the evolution of LLMs in desired directions.

For agents

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

get_harvested_code_for_paper("2404.04286")
get_code_for_paper("2404.04286")
have("2404.04286")

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

curl https://syntology.ai/api/ran/2404.04286.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.04286.svg)](https://syntology.ai/paper/2404.04286)

Connect an agent — have() is free.