SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.15449 · ICLR · 2025

Repetition Improves Language Model Embeddings

Graham Neubig, Aditi Raghunathan, Jacob Springer, Daniel Fried, Suhas Kotha

arXiv · PDF · Open in the Atlas

Code that ran

3 of 4 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 1 failure 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/2402.15449.json · badge: /api/ran/2402.15449.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
jakespringer/echo-embeddings canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
openbmb/debater community
Papers with Code archive
1 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
neuir/llm-qe community
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
EchoEmbeddingsMistral Ran
ran on a synthesized input · constructed an object; no result was computed
jakespringer/echo-embeddings/echo_embeddings.py
code served (permissive licence) · get_code("d764b2d25ca4c5ef")
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 an instance of its own class, EchoEmbeddingsMistral: the run constructed the object and did not call it, so no result was computed. 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 a minimal mock model that mimics AutoModel behavior
class MockTransformerModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.device = torch.device('cpu')
        self.linear = nn.Linear(4, 8)
    
    def forward(self, input_ids, attention_mask):
        # Simulate transformer output with last_hidden_state
        batch_size, seq_len = input_ids.shape
        hidden_dim = 8
        outputs = type('obj', (object,), {
            'last_hidden_state': self.linear(torch.randn(batch_size, seq_len, 4))
        })()
        return outputs

# Create the base model
base_model = MockTransformerModel()

# Create a minimal parser (can be None or a simple callable)
def mock_parser(x):
    return x

# Create a minimal pooling function (can be None or a simple callable)
def mock_pooling(x):
    return x.mean(dim=1)

# Construct the ARGS dictionary
ARGS = {
    'model': base_model,
    'parser': mock_parser,
    'pooling': mock_pooling,
}
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 EchoEmbeddingsMistral — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a EchoEmbeddingsMistral 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
LLM2Vec Ran
ran on a synthesized input · constructed an object; no result was computed
openbmb/debater/llm2vec/llm2vec.py
pointer only (licence: NONE) · get_code("54cf155a7282e094")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, openbmb/debater 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.

batch_to_device Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from mcgill-nlp/llm2vec
pointer only · get_code("74b280cb6a91117e")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. 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.

FlagModel Not yet run
unverified (harvested, no recorded run)
neuir/llm-qe/src/flagmodel.py
pointer only (licence: NONE) · get_code("8773b7cc3b3fff25")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, neuir/llm-qe 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. 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, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 1 later) and 3 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
batch_to_device
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 1 the same month, 1 later
2401.00368 (2024-01)
Improving Text Embeddings with Large Language Models
2402.09642 (2024-02)
Answer is All You Need: Instruction-following Text Embedding via Answering the Question
2502.12974 (2025-02)
Learning Refined Document Representations for Dense Retrieval via Deliberate Thinking

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

Bidirectional models are considered essential for strong text embeddings. Recent approaches to adapt autoregressive language models (LMs) into strong text embedding models have largely had the requirement to modify the LM architecture to be bidirectional. We challenge this premise by introducing "echo embeddings" which converts autoregressive LMs into high quality text embedding models without changing the architecture or requiring fine-tuning. By repeating the input and extracting embeddings from the repeated tokens-which have access to all original tokens-echo embeddings improve over classical LM embeddings by over 5% in zero-shot settings. Our zero-shot embeddings nearly match those obtained by bidirectionally-converted LMs that undergo additional masked-language modeling training. Echo embeddings are also compatible with supervised fine-tuning, matching or outperforming bidirectionally-converted LMs in an apples-to-apples comparison, even with an identical compute budget during training and inference. Overall, repetition is a simple and effective strategy to circumvent the need for bidirectional attention in embedding models, paving the way towards a unified architecture for all NLP tasks.

For agents

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

get_harvested_code_for_paper("2402.15449")
get_code_for_paper("2402.15449")
have("2402.15449")

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

curl https://syntology.ai/api/ran/2402.15449.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.15449.svg)](https://syntology.ai/paper/2402.15449)

Connect an agent — have() is free.