SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.14768 · NeurIPS · 2024

WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models

Peng Wang, Yong Jiang, Pengjun Xie, Fei Huang, Ningyu Zhang, Huajun Chen, Zexi Li, Yunzhi Yao, Ziwen Xu

arXiv · PDF · Open in the Atlas

Code that ran

13 of 16 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 3 have no contract verdict; 9 failures of our instrument; 3 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/2405.14768.json · badge: /api/ran/2405.14768.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
zjunlp/easyedit unrelated (classifier label)
classifier basis: classifier_2025
13 ran · 3 no recorded run
FunctionStatusWhere it lives
EditingMeanAct Ran
ran on a synthesized input
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("6556782be86e66b2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned NoneType. 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

class EditingMeanAct:
    """Computes and stores the average and current value"""

    def __init__(self, min_a=1000000000.0):
        self.reset(min_a=min_a)

    def reset(self, min_a=1000000000.0):
        self.avg = 0
        self.count = 0
        self.sum = 0
        self.min_a = min_a

    def update(self, val):
        self.sum += val
        self.count += 1
        self.avg = self.sum / self.count
        self.min_a = min(self.min_a, val)

    def mean_act(self):
        return self.avg

    def min_act(self):
        return self.min_a


INIT_ARGS = {
    'min_a': 1000000000.0
}

ARGS = {
    'val': 42.5
}

METHOD = 'update'

if __name__ == '__main__':
    instance = EditingMeanAct(**INIT_ARGS)
    instance.update(**ARGS)
    print(f"Mean: {instance.mean_act()}")
    print(f"Min: {instance.min_act()}")
    
    instance.update(val=10.0)
    print(f"Mean after second update: {instance.mean_act()}")
    print(f"Min after second update: {instance.min_act()}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type NoneType — a NoneType output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a NoneType 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
bernoulli Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("7e7272686fd1e3c2")
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 Tensor, ndim 2, dtype family float, shape (2, 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

torch.manual_seed(0)

# Create a small tensor with distinct dimensions for probing
tensor = torch.randn(2, 4, dtype=torch.float32)

# Define density as a float between 0 and 1
density = 0.7

# Define rescale as a boolean
rescale = True

ARGS = {
    "tensor": tensor,
    "density": density,
    "rescale": rescale,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
euc Ran
ran; our contract draft was wrong, not the code
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("e6c1e18615ecb8af")
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 1, dtype family float, shape (2,) 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
from transformers.activations import ACT2FN

torch.manual_seed(0)

# Create a minimal config class with hidden_act attribute
class MinimalConfig:
    def __init__(self):
        self.hidden_act = "gelu"  # Must be a valid key in ACT2FN

# Create tensors with distinct dimensions
# query and key should have shape (batch, seq_len, hidden_dim)
batch_size = 2
seq_len = 3
hidden_dim = 4

query = torch.randn(batch_size, seq_len, hidden_dim)
key = torch.randn(batch_size, seq_len, hidden_dim)

# Create config
config = MinimalConfig()

# Create act_mask with shape (batch, seq_len)
act_mask = torch.ones(batch_size, seq_len)

# Build ARGS dictionary
ARGS = {
    "query": query,
    "key": key,
    "config": config,
    "act_mask": act_mask,
    "infer": False
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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_mask Ran
ran; our contract draft was wrong, not the code
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("5bb8c583c306754b")
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 Tensor, ndim 2, dtype family float, shape (2, 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

torch.manual_seed(0)

# Create a small delta tensor with shape (batch=2, dim=4)
# This represents delta vectors from multiple models
delta = torch.randn(2, 4)

ARGS = {
    'delta': delta,
    'method': 'sum',
    'mask_dtype': torch.float32,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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_task_vectors Ran
ran; our contract draft was wrong, not the code
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("a0efe0cf5f74ca6c")
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

torch.manual_seed(0)

# Create a base tensor (small, CPU)
base = torch.randn(2, 4)

# Create a list of task tensors with the same shape as base
tensors = [
    torch.randn(2, 4),
    torch.randn(2, 4),
    torch.randn(2, 4),
]

ARGS = {
    "base": base,
    "tensors": tensors,
}
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
lerp Ran
ran, honoured the contract we drafted
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("1bfce29de3b973b3")
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 Tensor, ndim 2, dtype family float, shape (2, 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 numpy as np

torch.manual_seed(0)

# Create small test tensors and numpy arrays
t_value = 0.3
v0_tensor = torch.randn(2, 4)
v1_tensor = torch.randn(2, 4)
v0_numpy = np.random.randn(3, 5)
v1_numpy = np.random.randn(3, 5)

# Test with torch tensors
ARGS = {
    "t": 0.5,
    "v0": torch.randn(2, 4),
    "v1": torch.randn(2, 4)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
linear Ran
ran on a synthesized input
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("f3e855a894dfa058")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, 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

INIT_ARGS = {}

METHOD = 'execute'

ARGS = {
    't': 0.3,
    'v0': torch.randn(2, 4, 8),
    'v1': torch.randn(2, 4, 8),
    'DOT_THRESHOLD': 0.9995,
    'eps': 1e-08,
    'densities': None
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
magnitude Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("6a246076445216f3")
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 Tensor, ndim 2, dtype family float, shape (2, 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

torch.manual_seed(0)

# Create a small tensor with distinct dimensions
tensor = torch.randn(2, 4)

# Create a density value between 0 and 1
density = 0.5

ARGS = {
    "tensor": tensor,
    "density": density,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
maybe_torch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("10ff293597eab810")
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 Tensor, ndim 2, dtype family float, shape (2, 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 numpy as np

torch.manual_seed(0)

# Create a small numpy array
v = np.array([[1.0, 2.0, 3.0, 4.0],
              [5.0, 6.0, 7.0, 8.0]], dtype=np.float32)

# Test with is_torch=True
is_torch_true = True

# Test with is_torch=False
is_torch_false = False

ARGS = {
    "v": v,
    "is_torch": is_torch_true
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
normalize Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("a6e2f5aa10f37b10")
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 ndarray, ndim 1, dtype family float, shape (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 numpy as np

# Create a small vector to normalize
v = np.array([3.0, 4.0, 0.0, 0.0], dtype=np.float32)

# Define epsilon threshold
eps = 1e-6

ARGS = {
    "v": v,
    "eps": eps,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (4,) — shape (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
random_wo_rescaled Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("8b2a5f4da7fc7241")
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 Tensor, ndim 2, dtype family float, shape (2, 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

torch.manual_seed(0)

# Create a small tensor for testing
tensor = torch.randn(2, 4)

# Define density parameter
density = 0.5

# Construct ARGS dict with all parameters for random_wo_rescaled
ARGS = {
    "tensor": tensor,
    "density": density,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
rescaled_random Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("1abba92e438efda1")
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 Tensor, ndim 2, dtype family float, shape (2, 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

torch.manual_seed(0)

# Create a small tensor for testing
tensor = torch.randn(2, 4, dtype=torch.float32)

# Define density parameter
density = 0.7

# Create ARGS dict with all parameters for rescaled_random
ARGS = {
    "tensor": tensor,
    "density": density,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
slerp Ran
ran on a synthesized input
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("e595256c414955d3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, 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 numpy as np

METHOD = 'execute'

INIT_ARGS = {}

# Create two different vectors for interpolation
v0 = torch.randn(8)  # Starting vector
v1 = torch.randn(8)  # Ending vector

ARGS = {
    't': 0.5,  # Interpolation parameter (midpoint)
    'v0': v0,
    'v1': v1,
    'DOT_THRESHOLD': 0.9995,
    'eps': 1e-08,
    'densities': None
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
GTA Not yet run
unverified (harvested, no recorded run)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("a78f1fb529d93321")
How it ran

no invocation recorded.

WISE Not yet run
unverified (harvested, no recorded run)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("e9d558c06d4f86ba")
How it ran

no invocation recorded.

WISEAdapter Not yet run
unverified (harvested, no recorded run)
zjunlp/easyedit/easyeditor/models/wise/WISE.py
code served (permissive licence) · get_code("f978bb953aa03b08")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 later) and 14 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
lerp
brick of this paper's assembly: helper
1 other paper
1 later
2506.06292 (2025-06)
Mutual-Taught for Co-adapting Policy and Reward Models
maybe_torch
brick of this paper's assembly: helper
1 other paper
1 later
2506.06292 (2025-06)
Mutual-Taught for Co-adapting Policy and Reward Models

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

Large language models (LLMs) need knowledge updates to meet the ever-growing world facts and correct the hallucinated responses, facilitating the methods of lifelong model editing. Where the updated knowledge resides in memories is a fundamental question for model editing. In this paper, we find that editing either long-term memory (direct model parameters) or working memory (nonparametric knowledge of neural network activations/representations by retrieval) will result in an impossible triangle-reliability, generalization, and locality can not be realized together in the lifelong editing settings. For long-term memory, directly editing the parameters will cause conflicts with irrelevant pretrained knowledge or previous edits (poor reliability and locality). For working memory, retrieval-based activations can hardly make the model understand the edits and generalize (poor generalization). Therefore, we propose WISE to bridge the gap between memories. In WISE, we design a dual parametric memory scheme, which consists of the main memory for the pretrained knowledge and a side memory for the edited knowledge. We only edit the knowledge in the side memory and train a router to decide which memory to go through when given a query. For continual editing, we devise a knowledge-sharding mechanism where different sets of edits reside in distinct subspaces of parameters and are subsequently merged into a shared memory without conflicts. Extensive experiments show that WISE can outperform previous model editing methods and overcome the impossible triangle under lifelong model editing of question answering, hallucination, and out-of-distribution settings across trending LLM architectures, e.g., GPT, LLaMA, and Mistral ‡ .

For agents

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

get_harvested_code_for_paper("2405.14768")
get_code_for_paper("2405.14768")
have("2405.14768")

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

curl https://syntology.ai/api/ran/2405.14768.json

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

[![Syntology run record](https://syntology.ai/api/ran/2405.14768.svg)](https://syntology.ai/paper/2405.14768)

Connect an agent — have() is free.