SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2110.04374 · 2021

A Few More Examples May Be Worth Billions of Parameters

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 10 functions out of this paper's own repositories and ran 9 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
yuvalkirstain/lm-evaluation-harness canonical 9 of 10
FunctionStatusWhere it lives
mean Ran yuvalkirstain/lm-evaluation-harness/lm_eval/metrics.py
code served (permissive licence) · get_code("0928f497e20fb443")
get_result Ran yuvalkirstain/lm-evaluation-harness/lm_eval/models/gpt3.py
code served (permissive licence) · get_code("65ed12119df56f3b")
group Ran yuvalkirstain/lm-evaluation-harness/lm_eval/utils.py
code served (permissive licence) · get_code("359c35fdcd7a0ee5")
hash_args Ran yuvalkirstain/lm-evaluation-harness/lm_eval/base.py
code served (permissive licence) · get_code("ea06eaae4fc1eaf0")
is_number Ran yuvalkirstain/lm-evaluation-harness/lm_eval/utils.py
code served (permissive licence) · get_code("e18607dae38b30e2")
matthews_corrcoef Ran yuvalkirstain/lm-evaluation-harness/lm_eval/metrics.py
code served (permissive licence) · get_code("028eb6dad1493f40")
median Ran yuvalkirstain/lm-evaluation-harness/lm_eval/metrics.py
code served (permissive licence) · get_code("79a89132ce3f75f1")
simple_parse_args_string Ran yuvalkirstain/lm-evaluation-harness/lm_eval/utils.py
code served (permissive licence) · get_code("c6d952282e302ade")
write_datasets_fo_read Ran yuvalkirstain/lm-evaluation-harness/lm_eval/train_pl.py
code served (permissive licence) · get_code("31b87866d085212b")
load_raw_datasests Not yet run yuvalkirstain/lm-evaluation-harness/lm_eval/train_pl.py
code served (permissive licence) · get_code("3b0f3f81457fce8a")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We investigate the dynamics of increasing the number of model parameters versus the number of labeled examples across a wide variety of tasks. Our exploration reveals that while scaling parameters consistently yields performance improvements, the contribution of additional examples highly depends on the task's format. Specifically, in open question answering tasks, enlarging the training set does not improve performance. In contrast, classification, extractive question answering, and multiple choice tasks benefit so much from additional examples that collecting a few hundred examples is often "worth" billions of parameters. We hypothesize that unlike open question answering, which involves recalling specific information, solving strategies for tasks with a more restricted output space transfer across examples, and can therefore be learned with small amounts of labeled data.

For agents

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

get_harvested_code_for_paper("2110.04374")
get_code_for_paper("2110.04374")
have("2110.04374")

Connect an agent — have() is free.