SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2501.15225 · ACL · 2025

SEAL: Scaling to Emphasize Attention for Long-Context Retrieval

Eunhyeok Park, Minsang Seok, Changhun Lee, Jungyu Jin, Younghyun Cho

arXiv · PDF · Open in the Atlas

Code that ran

1 of 1 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; 1 failure of our instrument; 0 no recorded run), measured 2026-11-22; 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/2501.15225.json · badge: /api/ran/2501.15225.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
axolotl-ai-cloud/axolotl community
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
ssmax_flex_attention_forward Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
axolotl-ai-cloud/axolotl/src/axolotl/monkeypatch/scaled_softmax_attn.py
code served (permissive licence) · get_code("5eab3788d6530d2f")
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) 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)

# Mock the missing imports/functions that the function uses
class BlockMask:
    pass

def repeat_kv(x, num_reps):
    return x.repeat_interleave(num_reps, dim=1)

def compile_friendly_flex_attention(query, key, value, score_mod=None, block_mask=None, enable_gqa=True, scale=None, kernel_options=None, return_lse=False, training=False):
    # Mock implementation that returns attention output and optionally LSE
    batch_size, num_heads, seq_len, head_dim = query.shape
    attention_output = torch.randn(batch_size, num_heads, seq_len, head_dim)
    if return_lse:
        lse = torch.randn(batch_size, num_heads, seq_len)
        return attention_output, lse
    return attention_output

# Inject mocks into the module's namespace
import sys
current_module = sys.modules[__name__]
setattr(current_module, 'BlockMask', BlockMask)
setattr(current_module, 'repeat_kv', repeat_kv)
setattr(current_module, 'compile_friendly_flex_attention', compile_friendly_flex_attention)

# Create a minimal module
class SimpleModule(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x):
        return self.linear(x)

module = SimpleModule()
module.train()

# Create tensors with distinct dimensions
batch_size = 2
num_heads = 3
seq_len_q = 5
seq_len_kv = 6
head_dim = 4

query = torch.randn(batch_size, num_heads, seq_len_q, head_dim)
key = torch.randn(batch_size, num_heads, seq_len_kv, head_dim)
value = torch.randn(batch_size, num_heads, seq_len_kv, head_dim)

# Create attention mask (score_mask format)
attention_mask = torch.zeros(batch_size, 1, seq_len_q, seq_len_kv)

# Create position_ids
position_ids = torch.arange(seq_len_q).unsqueeze(0).expand(batch_size, -1)

ARGS = {
    'module': module,
    'query': query,
    'key': key,
    'value': value,
    'attention_mask': attention_mask,
    'scaling': 0.5,
    'softcap': None,
    'position_ids': position_ids,
    'dropout': 0.0,
    'kernel_options': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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

Also present in other papers' code

Of 1 samples lifted from this paper's linked repositories, 1 carry a function identity and 0 do not. Of the 1 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 1 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

While many advanced LLMs are designed to handle long sequence data, we can still observe notable quality degradation even within the sequence limit. In this work, we introduce a novel approach called Scaling to Emphasize Attention for Long-context retrieval (SEAL), which enhances the retrieval performance of large language models (LLMs) over long contexts. We observe that specific attention heads are closely tied to long-context retrieval, showing positive or negative correlation with retrieval scores, and adjusting the strength of these heads boosts the quality of LLMs in long context by a large margin. Built on this insight, we propose a learning-based mechanism that leverages generated data to emphasize these heads. By applying SEAL, we achieve significant improvements in long-context retrieval performance across various tasks and models. Additionally, when combined with existing training-free context extension techniques, SEAL extends the contextual limits of LLMs while maintaining highly reliable outputs.

For agents

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

get_harvested_code_for_paper("2501.15225")
get_code_for_paper("2501.15225")
have("2501.15225")

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

curl https://syntology.ai/api/ran/2501.15225.json

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

[![Syntology run record](https://syntology.ai/api/ran/2501.15225.svg)](https://syntology.ai/paper/2501.15225)

Connect an agent — have() is free.