SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.14641 · 2024

Distance between Relevant Information Pieces Causes Bias in Long-Context LLMs

arXiv · PDF · Open in the Atlas

Code that ran

4 of 5 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; 4 failures of our instrument; 1 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/2410.14641.json · badge: /api/ran/2410.14641.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
Rachum-thu/LongPiBench canonical (classifier label)
classifier basis: census
4 ran · 1 no recorded run
FunctionStatusWhere it lives
analyze_task_model Ran
ran; our contract draft was wrong, not the code
Rachum-thu/LongPiBench/visual/visualize.py
code served (permissive licence) · get_code("d29263a372e88f4b")
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 os
import json
import tempfile
from collections import defaultdict

# Create a temporary directory structure with test data
temp_dir = tempfile.mkdtemp()
res_dir = os.path.join(temp_dir, 'res')
model_dir = os.path.join(res_dir, 'test_model')
task_dir = os.path.join(model_dir, 'test_task')
os.makedirs(task_dir, exist_ok=True)

# Create a valid results.json file
results_data = [
    {'level': 'easy', 'score': 0.8},
    {'level': 'easy', 'score': 0.85},
    {'level': 'medium', 'score': 0.7},
    {'level': 'medium', 'score': 0.75},
    {'level': 'hard', 'score': 0.6},
    {'level': 'hard', 'score': 0.65},
]

results_path = os.path.join(task_dir, 'results.json')
with open(results_path, 'w') as f:
    json.dump(results_data, f)

# Change to the temporary directory so relative paths work
os.chdir(temp_dir)

ARGS = {
    'task': 'test_task',
    'model': 'test_model',
}
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 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
analyze_task_model Ran
ran; our contract draft was wrong, not the code
Rachum-thu/LongPiBench/visual/visualize_query.py
code served (permissive licence) · get_code("f92302698fc1cdeb")
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 os
import json
import tempfile
import shutil
from collections import defaultdict

# Create a temporary directory structure for test data
temp_dir = tempfile.mkdtemp()
res_base = os.path.join(temp_dir, 'res')
os.makedirs(res_base, exist_ok=True)

# Create a valid results.json file
model_name = "test_model"
task_name = "test_task"
model_dir = os.path.join(res_base, model_name, task_name)
os.makedirs(model_dir, exist_ok=True)

results_data = [
    {"level": "easy", "score": 0.95},
    {"level": "easy", "score": 0.92},
    {"level": "medium", "score": 0.85},
    {"level": "medium", "score": 0.88},
    {"level": "hard", "score": 0.75},
    {"level": "hard", "score": 0.78},
]

results_file = os.path.join(model_dir, 'results.json')
with open(results_file, 'w') as f:
    json.dump(results_data, f)

# Change working directory to temp_dir so relative paths work
original_cwd = os.getcwd()
os.chdir(temp_dir)

ARGS = {
    "task": "test_task",
    "model": "test_model",
}
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 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
darken_color Ran
ran; our contract draft was wrong, not the code
Rachum-thu/LongPiBench/visual/visualize.py
code served (permissive licence) · get_code("98a52295395bbc3a")
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 logging
import matplotlib.colors as mcolors

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

# Test case 1: hex color
color_hex = "#FF5733"
amount_1 = 0.7

# Test case 2: RGB tuple
color_rgb = (1.0, 0.5, 0.2)
amount_2 = 0.5

# Test case 3: named color
color_named = "red"
amount_3 = 0.3

ARGS = {
    "color": "#FF5733",
    "amount": 0.7
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib — 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
sanitize_filename Ran
ran; our contract draft was wrong, not the code
Rachum-thu/LongPiBench/visual/visualize.py
code served (permissive licence) · get_code("11bf077f72f9d4c0")
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 re

def sanitize_filename(name):
    """Replace unsafe characters in the string with underscores."""
    return re.sub('[^A-Za-z0-9_\\-]', '_', name)

ARGS = {
    'name': 'my-file_name@2024!.txt'
}
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 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
llm_generate Not yet run
unverified (harvested, no recorded run)
Rachum-thu/LongPiBench/src/llm/llm.py
code served (permissive licence) · get_code("e4196dba61412675")
How it ran

no invocation recorded.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 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 5 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

Positional bias in large language models (LLMs) hinders their ability to effectively process long inputs. A prominent example is the "lost in the middle" phenomenon, where LLMs struggle to utilize relevant information situated in the middle of the input. While prior research primarily focuses on single pieces of relevant information, real-world applications often involve multiple relevant information pieces. To bridge this gap, we present LongPiBench, a benchmark designed to assess positional bias involving multiple pieces of relevant information. Thorough experiments are conducted with five commercial and six open-source models. These experiments reveal that while most current models are robust against the "lost in the middle" issue, there exist significant biases related to the spacing of relevant information pieces. These findings highlight the importance of evaluating and reducing positional biases to advance LLM's capabilities.

For agents

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

get_harvested_code_for_paper("2410.14641")
get_code_for_paper("2410.14641")
have("2410.14641")

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

curl https://syntology.ai/api/ran/2410.14641.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.14641.svg)](https://syntology.ai/paper/2410.14641)

Connect an agent — have() is free.