Nicolas Zucchet, Antonio Orvieto
3 of 3 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 0 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/2405.21064.json · badge: /api/ran/2405.21064.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.
| Repository | Role | Ran |
|---|---|---|
| NicolasZucchet/Vanishing-and-exploding-gradients-are-not-the-end-of-the-story | canonical (classifier label) classifier basis: classifier_2025 |
2 ran · 0 no recorded run |
| copy not recorded | community not in the paper's declared repositories |
1 ran · 0 no recorded run |
| Function | Status | Where it lives |
|---|---|---|
| binary_operator_diag | Ran ran, honoured the contract we drafted |
this paper's copy was not recorded; identical code first harvested from nicolaszucchet/online-learning-lr-dependencies pointer only · get_code("0212f0ba3cfe6d69") How it ranThe 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. |
| matrix_init | Ran ran; our contract draft was wrong, not the code |
NicolasZucchet/Vanishing-and-exploding-gradients-are-not-the-end-of-the-story/source/model.py pointer only (licence: NONE) · get_code("76a18d8aa82c03b6") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, NicolasZucchet/Vanishing-and-exploding-gradients-are-not-the-end-of-the-story 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. |
| nu_init | Ran ran; our contract draft was wrong, not the code |
NicolasZucchet/Vanishing-and-exploding-gradients-are-not-the-end-of-the-story/source/model.py pointer only (licence: NONE) · get_code("4abcfdeb27acbd8a") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, NicolasZucchet/Vanishing-and-exploding-gradients-are-not-the-end-of-the-story 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. |
Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 2 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.
| Function | Also present in | Earliest-dated | Others named |
|---|---|---|---|
| binary_operator_diag | 3 other papers 3 dated earlier · 1 only as an assembly brick |
2111.00396 (2021-11) Efficiently Modeling Long Sequences with Structured State Spaces |
2305.15947 (2023-05) Online learning of long-range dependencies 2402.09900 (2024-02) Recurrent Reinforcement Learning with Memoroids |
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.
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Recurrent neural networks (RNNs) notoriously struggle to learn long-term memories, primarily due to vanishing and exploding gradients. The recent success of deep state-space models (SSMs), a subclass of RNNs, to overcome such difficulties challenges our theoretical understanding. In this paper, we delve into the optimization challenges of RNNs and discover that, as the memory of a network increases, changes in its parameters result in increasingly large output variations, making gradient-based learning highly sensitive, even without exploding gradients. Our analysis further reveals the importance of the element-wise recurrence design pattern combined with careful parametrizations in mitigating this effect. This feature is present in deep SSMs, as well as in other architectures, such as LSTMs. Overall, our insights provide a new explanation for some of the difficulties in gradient-based learning of RNNs and why some architectures perform better than others. Recurrent neural networks [RNNs; 1, 2] have long been the canonical architecture for modeling temporal data [3,4]. However, they are notoriously difficult to train on long sequences, as error signals flowing backward in time tend to either vanish or explode [5][6][7][8]. Attention mechanisms [9], as featured in transformers [10], address these issues by enabling direct token-to-token communication, considerably simplifying signal propagation across long time intervals. Yet, their superior performance comes with increased computational and memory costs, due to their quadratic scaling in the sequence length. This limitation has motivated significant research aimed at making transformers more efficient [11][12][13][14][15]. A promising line of research in this direction involves a new type of linear recurrent networks known as deep state-space models [SSMs; [16][17][18][19][20][21][22]. These models trade expressivity for faster training speed, and they have been shown to be particularly effective at capturing long-range dependencies. In this paper, we wonder whether this effectiveness can be solely attributed to their ability to avoid vanishing and exploding gradients. The simplicity of such models presents an opportunity for in-depth theoretical analysis. We focus on signal propagation within these models. After reviewing classical results on recurrent neural networks in Section 1, we demonstrate that they can suffer from an understudied problem: as the recurrent network encodes longer memories, the network's activity becomes increasingly sensitive to changes in its parameters, even when its dynamics remains stable. In Section 3, we then show that SSMs, as well as other architectures such as LSTMs, are well equipped to mitigate this issue. We then analyze a simple teacher-student task (Section 4). This task already reveals the remarkable complexity underlying the learning of linear recurrent networks and enables us to verify empirically our theory. Finally, we discuss how our findings extend to more realistic scenarios (Section 5), both in terms of architectures and data. Overall, our paper provides theoretical insights into the training of recurrent neural networks, an area where such analysis is rare. While vanishing and exploding gradients are well-known challenges, our results demonstrate that this is not the end of the story -there exists an additional layer of complexity beyond them. 38th Conference on Neural Information Processing Systems (NeurIPS 2024).
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2405.21064")
get_code_for_paper("2405.21064")
have("2405.21064")
The run record, dated, one paper per request, free:
curl https://syntology.ai/api/ran/2405.21064.json
A badge for a README (the split and the date, never a ratio):
[](https://syntology.ai/paper/2405.21064)
Connect an agent — have() is free.