SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2305.11539 · 2023

Delay-penalized CTC implemented based on Finite State Transducer

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 (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 0 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/2305.11539.json · badge: /api/ran/2305.11539.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
k2-fsa/k2 canonical (classifier label)
classifier basis: census
3 ran · 1 no recorded run
FunctionStatusWhere it lives
fix_for_boundary Ran
ran on a synthesized input
k2-fsa/k2/k2/python/k2/rnnt_loss.py
code served (permissive licence) · get_code("c220cdebd07ec5f3")
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 Tensor, ndim 3, dtype family float, shape (2, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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)

# Construct px: shape [B, S, T+1] where B=2, S=3, T+1=4
B, S, T1 = 2, 3, 4
px = torch.randn(B, S, T1, dtype=torch.float32)

# Construct boundary: shape [B, 3] containing [s_begin, t_begin, s_end, t_end]
# Only the 4th column (index 3, t_end) is used by the function
boundary = torch.tensor([
    [1, 0, 2, 2],  # batch 0: t_end = 2
    [1, 0, 2, 3],  # batch 1: t_end = 3
], dtype=torch.long)

ARGS = {
    "px": px,
    "boundary": boundary,
}
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 Tensor, ndim 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 4) as a later run recorded it
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
get_aux_label_info Ran
ran on a synthesized input
k2-fsa/k2/k2/python/k2/fsa.py
code served (permissive licence) · get_code("e73c35c6c9d0edd1")
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. 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
from typing import List, Optional, Tuple

torch.manual_seed(0)

# Test case 1: acceptor=True (should return 1 aux label)
ARGS_case1 = {
    "acceptor": True,
    "num_aux_labels": None,
    "aux_label_names": None,
}

# Test case 2: acceptor=False (should return 0 aux labels)
ARGS_case2 = {
    "acceptor": False,
    "num_aux_labels": None,
    "aux_label_names": None,
}

# Test case 3: aux_label_names provided
ARGS_case3 = {
    "acceptor": None,
    "num_aux_labels": None,
    "aux_label_names": ["aux_labels", "aux_labels2"],
}

# Test case 4: num_aux_labels provided
ARGS_case4 = {
    "acceptor": None,
    "num_aux_labels": 3,
    "aux_label_names": None,
}

# Test case 5: all None (should return 0 aux labels)
ARGS_case5 = {
    "acceptor": None,
    "num_aux_labels": None,
    "aux_label_names": None,
}

# Primary ARGS using case 3 as the main test case
ARGS = {
    "acceptor": None,
    "num_aux_labels": None,
    "aux_label_names": ["aux_labels", "aux_labels2"],
}
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 — 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
get_rnnt_logprobs Ran
ran on a synthesized input
k2-fsa/k2/k2/python/k2/rnnt_loss.py
code served (permissive licence) · get_code("2551b4f1c0679dda")
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. 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)

# Construct tensor arguments
B = 2  # batch size
S = 3  # symbol sequence length
T = 4  # acoustic sequence length
C = 5  # vocabulary size (including termination symbol)

# lm: Language model logprobs, shape [B][S+1][C]
lm = torch.randn(B, S + 1, C)

# am: Acoustic model logprobs, shape [B][T][C]
am = torch.randn(B, T, C)

# symbols: Symbol indices, shape [B][S]
symbols = torch.randint(0, C - 1, (B, S), dtype=torch.long)

# termination_symbol: integer in range [0, C-1]
termination_symbol = C - 1

# boundary: Optional tensor of shape [B, 4] with [begin_symbol, begin_frame, end_symbol, end_frame]
boundary = torch.tensor([
    [0, 0, S, T],
    [0, 0, S, T]
], dtype=torch.long)

# rnnt_type: string specifying the type
rnnt_type = "regular"

ARGS = {
    "lm": lm,
    "am": am,
    "symbols": symbols,
    "termination_symbol": termination_symbol,
    "rnnt_type": rnnt_type,
    "boundary": boundary
}
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 — 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
rnnt_loss_simple Not yet run
unverified (harvested, no recorded run)
k2-fsa/k2/k2/python/k2/rnnt_loss.py
code served (permissive licence) · get_code("fbd6cb5be5ae9a02")
How it ran

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, 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 4 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

Connectionist Temporal Classification (CTC) suffers from the latency problem when applied to streaming models. We argue that in CTC lattice, the alignments that can access more future context are preferred during training, thereby leading to higher symbol delay. In this work we propose the delay-penalized CTC which is augmented with latency penalty regularization. We devise a flexible and efficient implementation based on the differentiable Finite State Transducer (FST). Specifically, by attaching a binary attribute to CTC topology, we can locate the frames that firstly emit non-blank tokens on the resulting CTC lattice, and add the frame offsets to the log-probabilities. Experimental results demonstrate the effectiveness of our proposed delay-penalized CTC, which is able to balance the delay-accuracy trade-off. Furthermore, combining the delay-penalized transducer enables the CTC model to achieve better performance and lower latency. Our work is open-sourced and publicly available https://github.com/k2-fsa/k2.

For agents

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

get_harvested_code_for_paper("2305.11539")
get_code_for_paper("2305.11539")
have("2305.11539")

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

curl https://syntology.ai/api/ran/2305.11539.json

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

[![Syntology run record](https://syntology.ai/api/ran/2305.11539.svg)](https://syntology.ai/paper/2305.11539)

Connect an agent — have() is free.