SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2111.02570 · NeurIPS · 2021

CLUES: Few-Shot Learning Evaluation in Natural Language Understanding

Jianfeng Gao, Greg Yang, Hao Cheng, Xiaodong Liu, Subhabrata Mukherjee, Ahmed Awadallah, Guoqing Zheng, Saghar Hosseini, Christopher Meek

arXiv · PDF · Open in the Atlas

Code that ran

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; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 0 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/2111.02570.json · badge: /api/ran/2111.02570.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
microsoft/CLUES canonical (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
FunctionStatusWhere it lives
eval_S1 Ran
ran on a synthesized input
microsoft/CLUES/eval/S1.py
code served (permissive licence) · get_code("961d5ee547137316")
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 torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - eval_S1 takes pred and ans parameters
# - Both should be sequences (len() is called on them)
# - Each element pred[i] and ans[i] is passed to _s1_entry
# - _s1_entry treats its arguments as sequences (len() called, set() conversion)
# - _s1_entry computes intersection, so elements should be hashable

# The function expects pred and ans to be lists of lists (or sequences)
# where each inner list contains hashable items

# Create simple test data
pred = [
    [1, 2, 3],      # First prediction
    [4, 5],         # Second prediction
]

ans = [
    [1, 2, 4],      # First answer
    [4, 5, 6, 7],   # Second answer
]

ARGS = {
    'pred': pred,
    'ans': ans,
}
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 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
fill_labels Ran
ran on a synthesized input
microsoft/CLUES/classic_finetuning/prepro_clues_ext.py
code served (permissive licence) · get_code("72092c1103b26e43")
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
import numpy as np

torch.manual_seed(0)

# Construct arguments for fill_labels function
labels = ['O', 'O', 'O', 'O', 'O']
start = 1
end = 3
label_mapper = {
    'B-ENT': 0,
    'I-ENT': 1,
    'O': 2
}

ARGS = {
    'labels': labels,
    'start': start,
    'end': end,
    'label_mapper': label_mapper
}
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 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
search_index Ran
ran on a synthesized input
microsoft/CLUES/classic_finetuning/prepro_clues_ext.py
code served (permissive licence) · get_code("3dd493c26e93ff6c")
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 1, dtype family int; 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

torch.manual_seed(0)

# Construct valid arguments for search_index function
# The function searches for answer positions in tokenized text

# input_ids: list of token IDs
input_ids = [101, 2054, 2003, 1998, 2178, 102, 1045, 2572, 102]

# sequence_ids: list indicating which sequence each token belongs to
# 0 = query/first sequence, 1 = context/second sequence
# Using pad_on_right=False, so query tokens are marked as 1, context as 0
sequence_ids = [1, 1, 1, 1, 1, 0, 0, 0, 0]

# offsets: list of (start_char, end_char) tuples for each token
# Maps tokens back to character positions in original text
offsets = [
    (0, 0),      # [CLS]
    (1, 5),      # "What"
    (6, 10),     # "is"
    (11, 14),    # "and"
    (15, 20),    # "other"
    (21, 21),    # [SEP]
    (22, 26),    # "I"
    (27, 32),    # "answer"
    (33, 33),    # [SEP]
]

# cls_index: index of [CLS] token
cls_index = 0

# start_char: character position where answer starts in context
start_char = 27

# end_char: character position where answer ends in context
end_char = 33

# answer_in_query: whether the answer is in the query (False = answer in context)
answer_in_query = False

# pad_on_right: whether padding is on the right side
pad_on_right = False

ARGS = {
    "input_ids": input_ids,
    "sequence_ids": sequence_ids,
    "offsets": offsets,
    "cls_index": cls_index,
    "start_char": start_char,
    "end_char": end_char,
    "answer_in_query": answer_in_query,
    "pad_on_right": pad_on_right,
}
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, ndim 1, dtype family int — 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

Also present in other papers' code

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, 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 3 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

Most recent progress in natural language understanding (NLU) has been driven, in part, by benchmarks such as GLUE, SuperGLUE, SQuAD, etc. In fact, many NLU models have now matched or exceeded "human-level" performance on many tasks in these benchmarks. Most of these benchmarks, however, give models access to relatively large amounts of labeled data for training. As such, the models are provided far more data than required by humans to achieve strong performance. That has motivated a line of work that focuses on improving few-shot learning performance of NLU models. However, there is a lack of standardized evaluation benchmarks for few-shot NLU resulting in different experimental settings in different papers. To help accelerate this line of work, we introduce CLUES 1 , a benchmark for evaluating the few-shot learning capabilities of NLU models. We demonstrate that while recent models reach human performance when they have access to large amounts of labeled data, there is a huge gap in performance in the few-shot setting for most tasks. We also demonstrate differences between alternative model families and adaptation techniques in the few shot setting. Finally, we discuss several principles and choices in designing the experimental settings for evaluating the true few-shot learning performance and suggest a unified standardized approach to few-shot learning evaluation. We aim to encourage research on NLU models that can generalize to new tasks with a small number of examples. Code and data for CLUES are available at https://github.com/microsoft/CLUES.

For agents

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

get_harvested_code_for_paper("2111.02570")
get_code_for_paper("2111.02570")
have("2111.02570")

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

curl https://syntology.ai/api/ran/2111.02570.json

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

[![Syntology run record](https://syntology.ai/api/ran/2111.02570.svg)](https://syntology.ai/paper/2111.02570)

Connect an agent — have() is free.