SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2105.06232 · 2021

Retrieval-Free Knowledge-Grounded Dialogue Response Generation with Adapters

arXiv · PDF · Open in the Atlas

Code that ran

10 of 12 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 1 has no contract verdict; 8 failures of our instrument; 2 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/2105.06232.json · badge: /api/ran/2105.06232.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
hltchkust/knowexpert canonical (classifier label)
classifier basis: census
10 ran · 2 no recorded run
FunctionStatusWhere it lives
build_input_for_seq2seq_model Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/src/data_utils/utils.py
code served (permissive licence) · get_code("3d02a7d666422095")
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, 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

torch.manual_seed(0)

# The function expects:
# - labels: a tensor with at least 2 dimensions, where dim 1 is the sequence length
# - label_masks: a tensor with the same shape as labels
# The function uses .size(1) to get sequence length and .narrow() to slice along dimension 1

batch_size = 2
seqlen = 4

labels = torch.randint(0, 10, (batch_size, seqlen))
label_masks = torch.ones((batch_size, seqlen), dtype=torch.float32)

ARGS = {
    "labels": labels,
    "label_masks": label_masks,
}
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, ndim 3, dtype family float — a tuple output has no array shape of its own; none is stated
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
distinct_1 Ran
ran, honoured the contract we drafted
hltchkust/knowexpert/evaluation.py
code served (permissive licence) · get_code("721918dca3d398e7")
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 float, ndim 0, dtype family float, shape () 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 collections

def get_ngrams(text, n):
    """
    Returns all ngrams that are in the text.
    Note: this function does NOT lowercase text. If you want to lowercase, you should
    do so before calling this function.
    Inputs:
      text: string, space-separated
      n: int
    Returns:
      list of strings (each is a ngram, space-separated)
    """
    tokens = text.split()
    return [' '.join(tokens[i:i + n]) for i in range(len(tokens) - (n - 1))]

def get_ngram_counter(text, n):
    """
    Returns a counter, indicating how many times each n-gram appeared in text.
    Note: this function does NOT lowercase text. If you want to lowercase, you should
    do so before calling this function.
    Input:
      text: is a string, with tokens space-separated.
    Returns:
      counter: mapping from each n-gram (a space-separated string) appearing in text,
        to the number of times it appears
    """
    ngrams = get_ngrams(text, n)
    counter = collections.Counter()
    counter.update(ngrams)
    return counter

def _distinct_n(sample, n):
    """
    Returns (total number of unique ngrams in story_text) / (total number of ngrams in story_text, including duplicates).
    Text is lowercased before counting ngrams.
    Returns None if there are no ngrams
    """
    ngram_counter = get_ngram_counter(sample.strip().lower(), n)
    if sum(ngram_counter.values()) == 0:
        print('Warning: encountered a story with no {}-grams'.format(n))
        print(sample.strip().lower())
        print('ngram_counter: ', ngram_counter)
        return None
    return len(ngram_counter) / sum(ngram_counter.values())

def distinct_1(sample):
    return _distinct_n(sample, 1)


# Setup code to construct valid arguments
sample_text = "the quick brown fox jumps over the lazy dog the quick brown"

ARGS = {
    "sample": sample_text
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () 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
getDataLoader Ran
ran on a synthesized input
hltchkust/knowexpert/src/data_utils/data_reader.py
code served (permissive licence) · get_code("4cc91bb2f3ff1629")
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 DataLoader. 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
from torch.utils.data import TensorDataset

torch.manual_seed(0)

# Create a simple dataset with small tensors
# TensorDataset expects tensors and will return (x, y) tuples when iterated
x_data = torch.randn(2, 4)  # 2 samples, 4 features
y_data = torch.randint(0, 2, (2,))  # 2 labels
dataset = TensorDataset(x_data, y_data)

ARGS = {
    "dataset": dataset,
    "bsz": 2,
    "test": False
}
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 DataLoader — a DataLoader output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataLoader 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
get_data_from_batch Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/src/data_utils/dialog_reader.py
code served (permissive licence) · get_code("d8546d79a094a6b0")
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

torch.manual_seed(0)

# Create batch dictionary with required keys
batch_size = 2
seq_len = 4
context_dim = 3

batch = {
    'chosen_sentence': torch.randint(0, 100, (batch_size, seq_len)),
    'chosen_sentence_mask': torch.ones(batch_size, seq_len),
    'title': torch.randint(0, 100, (batch_size, seq_len)),
    'title_mask': torch.ones(batch_size, seq_len),
    'context': torch.randint(0, 100, (batch_size, seq_len)),
    'context_mask': torch.ones(batch_size, seq_len),
    'response': torch.randint(0, 100, (batch_size, seq_len)),
    'response_mask': torch.ones(batch_size, seq_len),
    'token_type': torch.randint(0, 2, (batch_size, seq_len)),
}

ARGS = {
    'batch': batch,
    'model_type': 'decoder_only'
}
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
get_ngram_counter Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/evaluation.py
code served (permissive licence) · get_code("b19a8662a8cd2ea8")
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 Counter. 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 collections

def get_ngrams(text, n):
    """
    Returns all ngrams that are in the text.
    Note: this function does NOT lowercase text. If you want to lowercase, you should
    do so before calling this function.
    Inputs:
      text: string, space-separated
      n: int
    Returns:
      list of strings (each is a ngram, space-separated)
    """
    tokens = text.split()
    return [' '.join(tokens[i:i + n]) for i in range(len(tokens) - (n - 1))]

def get_ngram_counter(text, n):
    """
    Returns a counter, indicating how many times each n-gram appeared in text.
    Note: this function does NOT lowercase text. If you want to lowercase, you should
    do so before calling this function.
    Input:
      text: is a string, with tokens space-separated.
    Returns:
      counter: mapping from each n-gram (a space-separated string) appearing in text,
        to the number of times it appears
    """
    ngrams = get_ngrams(text, n)
    counter = collections.Counter()
    counter.update(ngrams)
    return counter


ARGS = {
    "text": "the quick brown fox jumps over the lazy dog the quick fox",
    "n": 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Counter — a Counter output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Counter 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
get_ngrams Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/evaluation.py
code served (permissive licence) · get_code("8dced10689d4eef1")
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 list, 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

# Define ARGS dict with valid arguments for get_ngrams function
ARGS = {
    "text": "the quick brown fox jumps over the lazy dog",
    "n": 2
}
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 list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
load Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/src/data_utils/utils.py
code served (permissive licence) · get_code("1c4021c8a4208e22")
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 dict. 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 pickle
import tempfile
import os

# Create a temporary file with some pickled data
temp_dir = tempfile.gettempdir()
temp_filename = os.path.join(temp_dir, "test_pickle_load.pkl")

# Create test data and pickle it
test_data = {"key": "value", "number": 42, "list": [1, 2, 3]}
with open(temp_filename, 'wb') as f:
    pickle.dump(test_data, f)

ARGS = {
    'filename': temp_filename,
    'mode': 'rb'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
load_cmu_episodes Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/src/data_utils/cmu_dog_reader.py
code served (permissive licence) · get_code("eb1b7ecf101a0251")
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 list, 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 os
import tempfile
from pathlib import Path

# Create a temporary directory structure with the required files
temp_dir = tempfile.mkdtemp()
datasets_dir = os.path.join(temp_dir, 'datasets-CMU_DoG')
os.makedirs(datasets_dir, exist_ok=True)

# Create sample data files for 'train' mode
mode = 'train'

# Sample source data (context with 3 parts separated by < SEP >)
src_data = [
    "Hello < SEP > How are you < SEP > I'm fine",
    "What 's up < SEP > Not much < SEP > Just "chilling"",
]

# Sample target data (response)
tgt_data = [
    "I'm doing great!",
    "Nothing much here either.",
]

# Sample knowledge data (knowledge with 3 parts separated by < SEP >)
kn_data = [
    "fact1 < SEP > fact2 < SEP > fact3",
    "knowledge1 < SEP > knowledge2 < SEP > knowledge3",
]

# Write files
src_file = os.path.join(datasets_dir, f'src-{mode}-tokenized.txt')
tgt_file = os.path.join(datasets_dir, f'tgt-{mode}-tokenized.txt')
kn_file = os.path.join(datasets_dir, f'knl-{mode}-tokenized.txt')

with open(src_file, 'w') as f:
    f.write('\n'.join(src_data))

with open(tgt_file, 'w') as f:
    f.write('\n'.join(tgt_data))

with open(kn_file, 'w') as f:
    f.write('\n'.join(kn_data))

ARGS = {
    'data_dir': temp_dir,
    'mode': 'train',
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
pad_sents Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/src/data_utils/utils.py
code served (permissive licence) · get_code("71dac7a11a85c231")
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)

# sents parameter: a list of sentences, where each sentence is a list of integers (token ids)
# We'll create a few sentences of different lengths
sents = [
    [1, 2, 3, 4, 5],
    [10, 11, 12],
    [20, 21, 22, 23, 24, 25, 26]
]

ARGS = {
    'sents': sents,
    'pad_token': 0,
    'max_len': 512
}
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
preprocess_wow_episodes Ran
ran; our contract draft was wrong, not the code
hltchkust/knowexpert/src/data_utils/data_reader.py
code served (permissive licence) · get_code("169dee0029e8d444")
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 list, 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 math
from tqdm import tqdm

def preprocess_wow_episodes(episodes, mode, history_in_context, max_episode_length=1, cal_time=False):
    new_episodes = []
    for episode_num, episode in enumerate(tqdm(episodes, desc='Preprocess episodes', ncols=100)):
        for example_num, example in enumerate(episode):
            new_examples = {'context': [], 'response': '', 'title': '', 'topic': '', 'checked_sentence': '', 'knowledge': [], 'sample_id': example_num, 'episode_num': episode_num}
            history = []
            if example_num != 0 and history_in_context:
                max_pair_length = math.ceil(max_episode_length / 2)
                start_idx = max(0, example_num - max_pair_length)
                for num in range(start_idx, example_num):
                    if max_episode_length % 2 == 0 or num != start_idx:
                        history.append(episode[num]['text'].lower().strip())
                    history.append(episode[num]['labels'][0].lower().strip() if mode == 'train' else episode[num]['eval_labels'][0].lower().strip())
                assert len(history) == max_episode_length
            context = history + [example['text'].lower().strip()]
            if mode == 'train':
                response = example['labels'][0]
            else:
                response = example['eval_labels'][0]
            title = example['title']
            checked_sentence = example['checked_sentence']
            topic = example['chosen_topic']
            knowledges = example['knowledge'].rstrip().split('\n')
            new_examples['context'] = context
            new_examples['response'] = response.lower()
            new_examples['title'] = title.lower()
            new_examples['topic'] = topic
            new_examples['checked_sentence'] = checked_sentence.lower()
            new_examples['knowledge'] = knowledges
            new_episodes.append(new_examples)
    if 'test' in mode and cal_time:
        print('Testing the efficiency of the model!')
        with open(f'Efficiency/history_{mode}_ids.json', 'r') as f:
            filter_ids = json.load(f)
        temp = []
        for idx in range(len(new_episodes)):
            if idx in filter_ids:
                temp.append(new_episodes[idx])
        new_episodes = temp
        print(f'There are {len(new_episodes)} samples in the final testing list.')
    return new_episodes


example_train = {
    'text': 'Hello there',
    'labels': ['Hi how are you'],
    'eval_labels': ['Hi there'],
    'title': 'Greeting',
    'checked_sentence': 'A polite greeting',
    'chosen_topic': 'conversation',
    'knowledge': 'fact1\nfact2\nfact3'
}

example_train_2 = {
    'text': 'How are you doing',
    'labels': ['I am doing great'],
    'eval_labels': ['Doing well'],
    'title': 'Status',
    'checked_sentence': 'Asking about status',
    'chosen_topic': 'health',
    'knowledge': 'fact4\nfact5'
}

example_test = {
    'text': 'What is your name',
    'labels': ['My name is Bot'],
    'eval_labels': ['I am an AI'],
    'title': 'Identity',
    'checked_sentence': 'Asking for identity',
    'chosen_topic': 'identity',
    'knowledge': 'fact6'
}

episode_train = [example_train, example_train_2]
episode_test = [example_test]

episodes_train = [episode_train]
episodes_test = [episode_test]

ARGS = {
    'episodes': episodes_train,
    'mode': 'train',
    'history_in_context': False,
    'max_episode_length': 1,
    'cal_time': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
prediction_step Not yet run
unverified (harvested, no recorded run)
hltchkust/knowexpert/inference_task.py
code served (permissive licence) · get_code("5c4e4584b5aaf31c")
How it ran

no invocation recorded.

prepare_inputs_for_generation Not yet run
unverified (harvested, no recorded run)
hltchkust/knowexpert/inference_task.py
code served (permissive licence) · get_code("67735cd057723d31")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 11 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
getDataLoader 1 other paper
1 later
2106.03530 (2021-06)
CAiRE in DialDoc21: Data Augmentation for Information-Seeking Dialogue System

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

To diversify and enrich generated dialogue responses, knowledge-grounded dialogue has been investigated in recent years. The existing methods tackle the knowledge grounding challenge by retrieving the relevant sentences over a large corpus and augmenting the dialogues with explicit extra information. Despite their success, however, the existing works have drawbacks in inference efficiency. This paper proposes KnowExpert, a framework to bypass the explicit retrieval process and inject knowledge into the pre-trained language models with lightweight adapters and adapt to the knowledge-grounded dialogue task. To the best of our knowledge, this is the first attempt to tackle this challenge without retrieval in this task under an open-domain chit-chat scenario. The experimental results show that Knowexpert performs comparably with some retrieval-based baselines while being time-efficient in inference, demonstrating the effectiveness of our proposed method.

For agents

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

get_harvested_code_for_paper("2105.06232")
get_code_for_paper("2105.06232")
have("2105.06232")

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

curl https://syntology.ai/api/ran/2105.06232.json

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

[![Syntology run record](https://syntology.ai/api/ran/2105.06232.svg)](https://syntology.ai/paper/2105.06232)

Connect an agent — have() is free.