SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.11745 · 2024

FocusLLM: Precise Understanding of Long Context by Dynamic Condensing

arXiv · PDF · Open in the Atlas

Code that ran

12 of 13 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; 12 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/2408.11745.json · badge: /api/ran/2408.11745.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
leezythu/focusllm canonical (classifier label)
classifier basis: census
12 ran · 1 no recorded run
FunctionStatusWhere it lives
chunk_generate Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/eval_yarn_mistral.py
code served (permissive licence) · get_code("62f06040219b1d48")
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
import torch.nn as nn
from transformers.modeling_outputs import BaseModelOutputWithPast

torch.manual_seed(0)

# Stand-in tokenizer
class StandInTokenizer:
    def __init__(self):
        self.pad_token_id = 0
    
    def __call__(self, texts, return_tensors=None, padding=None):
        # Return a mock tokenizer output
        batch_size = len(texts)
        seq_len = 5
        input_ids = torch.randint(1, 100, (batch_size, seq_len))
        attention_mask = torch.ones(batch_size, seq_len)
        
        class TokenizerOutput:
            def to(self, device):
                self.input_ids = self.input_ids.to(device)
                self.attention_mask = self.attention_mask.to(device)
                return self
        
        output = TokenizerOutput()
        output.input_ids = input_ids
        output.attention_mask = attention_mask
        return output
    
    def decode(self, ids, skip_special_tokens=False):
        return "generated text"

# Stand-in model with minimal structure
class StandInModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.device = torch.device('cpu')
        self.model = self
        self.num_layers = 2
        self.hidden_size = 4
        self.num_heads = 2
    
    def forward(self, input_ids, attention_mask, position_ids, past_key_values=None, return_dict=False, use_cache=False):
        batch_size = input_ids.shape[0]
        seq_len = input_ids.shape[1]
        
        # Create dummy hidden states
        hidden_states = torch.randn(batch_size, seq_len, self.hidden_size)
        
        # Create dummy KV cache
        new_cache = ()
        for layer_i in range(self.num_layers):
            # Keys and values have shape (batch, num_heads, seq_len, head_dim)
            total_seq_len = seq_len if past_key_values is None else (past_key_values[layer_i][0].shape[2] + seq_len)
            keys = torch.randn(batch_size, self.num_heads, total_seq_len, self.hidden_size // self.num_heads)
            values = torch.randn(batch_size, self.num_heads, total_seq_len, self.hidden_size // self.num_heads)
            new_cache += ((keys, values),)
        
        output = BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=new_cache
        )
        return output
    
    def generate(self, input_ids, attention_mask, max_new_tokens, past_key_values, eos_token_id, use_cache):
        # Mock generate: return input_ids with appended tokens
        batch_size = input_ids.shape[0]
        new_tokens = torch.randint(1, 100, (batch_size, max_new_tokens))
        output = torch.cat([input_ids, new_tokens], dim=1)
        return output

# Create instances
tok = StandInTokenizer()
model = StandInModel()
texts = ["hello world", "foo bar"]
max_tokens = 5
sliding_window = 128 * 1024
chunk_size = 2500
verbose = False

ARGS = {
    "model": model,
    "tok": tok,
    "texts": texts,
    "max_tokens": max_tokens,
    "sliding_window": sliding_window,
    "chunk_size": chunk_size,
    "verbose": verbose,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — 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
create_system_msg Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/eval_utils.py
code served (permissive licence) · get_code("ab36ebf61cfb78fb")
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
import torch.nn as nn

torch.manual_seed(0)

# The function only takes a data_name parameter which is a string
# It branches on the value and returns a system message string
# No complex objects are needed

ARGS = {
    'data_name': 'math_calc'
}
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
f1_score Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/compute_scores.py
code served (permissive licence) · get_code("43daed1a27177d83")
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 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.

from collections import Counter

def f1_score(prediction, ground_truth) -> tuple[float, float, float]:
    common = Counter(prediction) & Counter(ground_truth)
    num_same = sum(common.values())
    if num_same == 0:
        return (0, 0, 0)
    precision = 1.0 * num_same / len(prediction)
    recall = 1.0 * num_same / len(ground_truth)
    f1 = 2 * precision * recall / (precision + recall)
    return (f1, precision, recall)


ARGS = {
    "prediction": [1, 2, 3, 2, 1, 4],
    "ground_truth": [1, 2, 2, 3, 5, 1]
}
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 tuple, ndim 1, dtype family float — 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
format_numel_str Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/src/colossal.py
code served (permissive licence) · get_code("57207a94cde43fcb")
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)

# The function format_numel_str takes a single integer parameter 'numel'
# and returns a formatted string representation of that number in terms of
# billions (B), millions (M), thousands (K), or as-is.

# Test with a value that exercises the function logic
# Using a number in the millions range
numel_value = 5242880  # This is 5 * 1024^2, should format as "5.00 M"

ARGS = {
    'numel': numel_value
}
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
generate_prompt_landmark Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/eval_passkey.py
code served (permissive licence) · get_code("1cea3f6235148922")
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 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.

from numpy import random

ARGS = {
    'n_garbage': 1000,
    'seed': 42
}
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 tuple, ndim 1, dtype family float — 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
load_data Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/eval_utils.py
code served (permissive licence) · get_code("ded262d73a687b82")
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 json
import tempfile
from pathlib import Path

# Create a temporary directory and a sample JSONL file
temp_dir = tempfile.mkdtemp()
data_dir = temp_dir

# Create a sample passkey.jsonl file
sample_data = [
    {"text": "The answer is 42", "answer": "42"},
    {"text": "Find the number: 123", "answer": "123"},
    {"text": "Key information here", "answer": "key"},
]

passkey_file = Path(data_dir, "passkey.jsonl")
with open(passkey_file, 'w') as f:
    for item in sample_data:
        f.write(json.dumps(item) + "\n")

# Define ARGS with all parameters for load_data function
ARGS = {
    "data_name": "passkey",
    "data_dir": data_dir,
}
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, 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
load_json Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/eval_utils.py
code served (permissive licence) · get_code("41e438356614fa31")
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 json
import tempfile
import os

# Create a temporary JSON file for testing
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test.json")

# Write a simple JSON file
test_data = {"key": "value", "number": 42, "list": [1, 2, 3]}
with open(temp_file, 'w') as f:
    json.dump(test_data, f)

ARGS = {
    "fname": temp_file
}
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
load_json Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/src/colossal.py
code served (permissive licence) · get_code("9edde0325b62d764")
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 json
import tempfile
import os

# Create a temporary JSON file with some valid data
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test.json")

test_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "object"}
}

with open(test_file, 'w') as f:
    json.dump(test_data, f)

ARGS = {
    "file_path": test_file
}
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
normalize_answer Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/compute_scores.py
code served (permissive licence) · get_code("4b44a4d3fd80a48d")
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
import string

def normalize_answer(s: str) -> str:
    """Lower text and remove punctuation, articles and extra whitespace."""

    def remove_articles(text):
        return re.sub('\\b(a|an|the)\\b', ' ', text)

    def white_space_fix(text):
        return ' '.join(text.split())

    def remove_punc(text):
        exclude = set(string.punctuation)
        return ''.join((ch for ch in text if ch not in exclude))

    def lower(text):
        return text.lower()
    return white_space_fix(remove_articles(remove_punc(lower(s))))


ARGS = {
    "s": "The Quick Brown Fox! How are you, today?"
}
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
normalize_zh_answer Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/compute_scores.py
code served (permissive licence) · get_code("f62b4d2a31aa1081")
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 string

def normalize_zh_answer(s: str) -> str:
    """Chinese version. Lower text and remove punctuation, extra whitespace."""

    def white_space_fix(text):
        return ''.join(text.split())

    def remove_punc(text):
        cn_punctuation = '!?。。"#$%&'()*+,-/:;<=>@[\]^_`{|}~⦅⦆「」、、〃》「」『』【】〔〕〖〗〘〙〚〛〜〝〞〟〰〾〿–—''‛""„‟…‧﹏.'
        all_punctuation = set(string.punctuation + cn_punctuation)
        return ''.join((ch for ch in text if ch not in all_punctuation))

    def lower(text):
        return text.lower()
    return white_space_fix(remove_punc(lower(s)))


# Test string with Chinese and English characters, punctuation, and whitespace
test_string = "Hello  世界!?This is a Test。with、mixed,punctuation"

ARGS = {
    "s": test_string
}
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
truncate_by_tokens Ran
ran; our contract draft was wrong, not the code
leezythu/focusllm/infbench_src/eval_chatglm.py
pointer only (licence: MIT); the stored copy get_code serves, from franxyao/long-context-data-engineering, is pointer only (licence: NONE) · get_code("90a4b8ca09aaaf67")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from franxyao/long-context-data-engineering, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

truncate_input Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
leezythu/focusllm/infbench_src/eval_chatglm.py
pointer only (licence: MIT); the stored copy get_code serves, from franxyao/long-context-data-engineering, is pointer only (licence: NONE) · get_code("a09f1fe8cdcea888")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from franxyao/long-context-data-engineering, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

passkey_retrieval_test Not yet run
unverified (harvested, no recorded run)
leezythu/focusllm/eval_passkey.py
code served (permissive licence) · get_code("0f44a5e49cd59e9f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 13 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 12 later) and 0 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.

FunctionAlso present inEarliest-datedOthers named
normalize_answer 5 other papers
3 dated earlier, 2 later
2402.04617 (2024-02)
InfLLM: Training-Free Long-Context Extrapolation for LLMs with an Efficient Context Memory
2406.07528 (2024-06)
QuickLLaMA: Query-aware Inference Acceleration for Large Language Models
2407.09450 (2024-07)
Human-inspired Episodic Memory for Infinite Context LLMs
and 2 more not named here
normalize_zh_answer 4 other papers
3 dated earlier, 1 later
2402.04617 (2024-02)
InfLLM: Training-Free Long-Context Extrapolation for LLMs with an Efficient Context Memory
2406.07528 (2024-06)
QuickLLaMA: Query-aware Inference Acceleration for Large Language Models
2407.09450 (2024-07)
Human-inspired Episodic Memory for Infinite Context LLMs
and 1 more not named here
truncate_by_tokens 3 other papers
2 dated earlier, 1 later
2402.10171 (2024-02)
Data Engineering for Scaling Language Models to 128K Context
2402.13718 (2024-02)
$\infty$Bench: Extending Long Context Evaluation Beyond 100K Tokens
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
truncate_input 3 other papers
2 dated earlier, 1 later
2402.10171 (2024-02)
Data Engineering for Scaling Language Models to 128K Context
2402.13718 (2024-02)
$\infty$Bench: Extending Long Context Evaluation Beyond 100K Tokens
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
chunk_generate 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
create_system_msg 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
f1_score 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
format_numel_str 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
generate_prompt_landmark 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
load_data 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
load_json 1 other paper
1 dated earlier
2310.05737 (2023-10)
Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation
load_json 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing
passkey_retrieval_test 1 other paper
1 later
2025.acl-long.1500 (2025 (year only))
FocusLLM: Precise Understanding of Long Context by Dynamic Condensing

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

Empowering LLMs with the ability to precisely understand long contexts is crucial for many downstream applications. However, handling long contexts with conventional transformer architecture requires substantial training and inference resources. Existing context condensing methods cannot accurately understand the full context, as there is a considerable amount of information loss in the condensing process. To address these issues, we present FocusLLM, a framework designed to extend the fixed context length of any decoder-only LLM, allowing the model to focus on relevant information from very long sequences. FocusLLM first divides long text input into chunks based on the model's original context length. It then employs the dynamic condensing process to distill crucial information from each chunk. Ultimately, through the novel parallel decoding mechanism, FocusLLM can integrate the extracted information into its local context. FocusLLM stands out for great training efficiency and versatility: trained with an 8K input length and with much less training cost than previous methods, FocusLLM exhibits superior performance across downstream tasks and maintains strong language modeling ability when handling extensive long texts, even up to 400K tokens. Our code is available at https://github.com/leezythu/FocusLLM.

For agents

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

get_harvested_code_for_paper("2408.11745")
get_code_for_paper("2408.11745")
have("2408.11745")

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

curl https://syntology.ai/api/ran/2408.11745.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.11745.svg)](https://syntology.ai/paper/2408.11745)

Connect an agent — have() is free.