SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.26690 · 2026

Self-Improvement Imitation with Biologically Guided Search for Protein Design Under Oracle Budgets

Ashima Khanna, Dominik Grimm

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 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; 4 failures of our instrument; 2 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/2605.26690.json · badge: /api/ran/2605.26690.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
samsinai/FLEXS canonical (classifier label)
classifier basis: census
1 ran · 0 no recorded run
hyeonahkimm/delta_cs community
Papers with Code archive
2 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 0 no recorded run
HySonLab/Directed_Evolution community
Papers with Code archive
1 ran · 2 no recorded run
FunctionStatusWhere it lives
MbStack Ran
ran on a synthesized input · constructed an object; no result was computed
hyeonahkimm/delta_cs/BioSeq-GFN-AL/run_amp_delta.py
code served (permissive licence) · get_code("50b2344935528dab")
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, MbStack: 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)

# Define a simple function that takes a list of tensors and returns a list of tensors
def simple_f(batch):
    """Takes a list of tensors and returns a list of processed tensors"""
    return [x.sum(dim=1, keepdim=True) for x in batch]

ARGS = {
    "f": simple_f
}
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 MbStack — 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 MbStack 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
RolloutWorker Ran
ran on a synthesized input · constructed an object; no result was computed
hyeonahkimm/delta_cs/BioSeq-GFN-AL/run_amp_delta.py
code served (permissive licence) · get_code("87be593b7402674f")
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, RolloutWorker: 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
import numpy as np

torch.manual_seed(0)

class MockTokenizer:
    def __init__(self):
        self.eos_token = '<eos>'
        self.vocab = MockVocab()
    
    def process(self, texts):
        # Return a dummy tensor of shape (batch_size, seq_len)
        batch_size = len(texts)
        return torch.randn(batch_size, 5)

class MockVocab:
    def __init__(self):
        self.itos = {0: 'a', 1: 'b', 2: 'c', 3: '<eos>', 4: 'd'}

class MockOracle(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(5, 1)
    
    def forward(self, x_list):
        # x_list is a list of tensors
        results = []
        for x in x_list:
            if isinstance(x, str):
                results.append(0.5)
            else:
                results.append(0.5)
        return results

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.model = nn.Sequential(
            nn.Linear(5, 8),
            nn.ReLU(),
            nn.Linear(8, 10)  # 10 vocab size
        )
    
    def forward(self, x, lens, coef=1.0, pad=22):
        # x shape: (batch_size, seq_len)
        # Return logits of shape (batch_size, vocab_size)
        batch_size = x.shape[0]
        return self.model(x[:, :5])

class MockArgs:
    def __init__(self):
        self.max_len = 8
        self.gen_max_len = 10
        self.gen_episodes_per_step = 2
        self.gen_random_action_prob = 0.1
        self.gen_reward_exp = 1.0
        self.gen_sampling_temperature = 1.0
        self.gen_output_coef = 1.0
        self.gen_balanced_loss = 0
        self.gen_reward_norm = 1.0
        self.gen_reward_min = 0.0
        self.gen_loss_eps = 1e-8
        self.device = 'cpu'
        self.gen_leaf_coef = 0.1
        self.gen_reward_exp_ramping = 0.0

oracle = MockOracle()
tokenizer = MockTokenizer()
args = MockArgs()

ARGS = {
    'args': args,
    'oracle': oracle,
    'tokenizer': tokenizer
}
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 RolloutWorker — 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 RolloutWorker 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
filter_len Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from mj10/bioseq-gfn-al
pointer only · get_code("a0962a1a5631fcc1")
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.

get_current_radius Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from hyeonahkimm/delta_cs
pointer only · get_code("7a638b085246b050")
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.

r2_weights Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
samsinai/FLEXS/flexs/baselines/models/adaptive_ensemble.py
code served (permissive licence) · get_code("48932a6193ef2693")
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 ndarray, ndim 1, dtype family float, shape (3,) 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
import scipy.stats

# Create sample data
num_models = 3
num_samples = 5

# Generate model predictions: shape (num_models, num_samples)
np.random.seed(0)
model_preds = np.random.randn(num_models, num_samples)

# Generate labels: shape (num_samples,)
labels = np.random.randn(num_samples)

ARGS = {
    "model_preds": model_preds,
    "labels": labels,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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 (3,) — shape (3,) 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
timer Ran
ran; our contract draft was wrong, not the code
HySonLab/Directed_Evolution/de/directed_evolution.py
pointer only (licence: GPL-3.0) · get_code("883df0e10ee73f1c")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, HySonLab/Directed_Evolution, is not in the permissive set Syntology serves inline. 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.

BaseMasker Not yet run
unverified (harvested, no recorded run)
HySonLab/Directed_Evolution/de/directed_evolution.py
pointer only (licence: GPL-3.0) · get_code("dbcfa1ea15abeec5")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, HySonLab/Directed_Evolution, is not in the permissive set Syntology serves inline. 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.

DiscreteDirectedEvolution2 Not yet run
unverified (harvested, no recorded run)
HySonLab/Directed_Evolution/de/directed_evolution.py
pointer only (licence: GPL-3.0) · get_code("d59873ee5b23e4b1")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, HySonLab/Directed_Evolution, is not in the permissive set Syntology serves inline. 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 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 6 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
filter_len
brick of this paper's assembly: helper
3 other papers
3 dated earlier
2203.04115 (2022-03)
Biological Sequence Design with GFlowNets
2302.09465 (2023-02)
Stochastic Generative Flow Networks
2410.04461 (2024-10)
Improved Off-policy Reinforcement Learning in Biological Sequence Design
get_current_radius
brick of this paper's assembly: scheduler
1 other paper
1 dated earlier
2410.04461 (2024-10)
Improved Off-policy Reinforcement Learning in Biological Sequence Design

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

Protein sequence optimization under tight oracle budgets requires methods that explore vast combinatorial spaces while making each evaluation informative. Existing reinforcement learning and off-policy generative approaches often degrade under surrogate noise, and position-agnostic mutation proposals risk disrupting functionally critical residues. We introduce SILO, a trajectory-level self-improvement imitation framework for oracle-budgeted protein design. SILO uses a hierarchical edit policy that decomposes each mutation into a position choice followed by a residue choice. In each active-learning round, the policy samples candidate trajectories via incremental stochastic beam search without replacement (SBS), and a UCB-based proxy ensemble, combined with an alanine-scan fitness score (AFS), selects candidates with functionally relevant edits for in silico oracle evaluation. The policy is then updated by next-action cross-entropy imitation on the round's best oracle-labeled trajectories, avoiding value-function estimation. Across eight reproduced protein fitness landscapes and five strong baselines from prior work, SILO achieves the highest maximum and top-100 mean fitness on 8 of 8 landscapes within our evaluations, often exhibiting faster early-stage improvement. In low-data and noisy-proxy stress tests on two landscapes per setting, SILO remains competitive or best when several baselines degrade. Ablations show that SBS with AFS account for much of the gains, with iterative imitation providing additional improvement.

For agents

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

get_harvested_code_for_paper("2605.26690")
get_code_for_paper("2605.26690")
have("2605.26690")

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

curl https://syntology.ai/api/ran/2605.26690.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.26690.svg)](https://syntology.ai/paper/2605.26690)

Connect an agent — have() is free.