SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.15011 · EMNLP Findings · 2020

TLDR: Extreme Summarization of Scientific Documents

Arman Cohan, Kyle Lo, Daniel Weld, Isabel Cachola

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 19 functions out of this paper's own repositories and ran 17 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
allenai/scitldr canonical 5 of 5
bnitsan/papertweet — 9 of 10
jungokasai/twist_decoding — 3 of 4
FunctionStatusWhere it lives
CustomDataset Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("c1e87177e1879e25")
add_ctrl Ran allenai/scitldr/SciTLDR-Data/build_ctrl_datasets.py
code served (permissive licence) · get_code("34369269378d0047")
add_eos_token_to_target Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("5a617f87118cdcb7")
add_prompt_to_source Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("a6e6d89d8756add2")
build_batches Ran allenai/scitldr/SciTLDR-Data/to_stories.py
code served (permissive licence) · get_code("c62c9d8fca3d91e9")
get_line_ids Ran jungokasai/twist_decoding/fairseq/twist/generate_twist_tldr.py
pointer only (licence: NONE) · get_code("4f101059eeab11c8")
get_num_train_steps Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("39d9cd45b8edce68")
get_train_dataloaders Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("cb94a51b64280335")
get_url_hashes Ran allenai/scitldr/SciTLDR-Data/make_datafiles.py
code served (permissive licence) · get_code("478f0055455e2023")
get_val_dataloaders Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("238977a0b4ed33b4")
hashhex Ran allenai/scitldr/SciTLDR-Data/make_datafiles.py
code served (permissive licence) · get_code("2ed659a3dd835939")
jaccard Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("18e9681c72c82d0d")
pad_candidates Ran jungokasai/twist_decoding/fairseq/twist/generate_twist_tldr.py
pointer only (licence: NONE) · get_code("258d4a4b0a8b6b16")
read_input Ran jungokasai/twist_decoding/fairseq/twist/generate_twist_tldr.py
pointer only (licence: NONE) · get_code("53d5b85264377a72")
read_text_file Ran allenai/scitldr/SciTLDR-Data/make_datafiles.py
code served (permissive licence) · get_code("8175760b8eddec2b")
rouge_score Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("9447eeb9dbba8819")
train_val_test_path Ran bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("f6ceffd4a0becc2a")
CondGenModel Not yet run bnitsan/papertweet/models/model.py
code served (permissive licence) · get_code("0e0ae26c9368468f")
generate_TLDRs Not yet run jungokasai/twist_decoding/fairseq/twist/generate_twist_tldr.py
pointer only (licence: NONE) · get_code("bba331b9f86f6008")

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 introduce TLDR generation, a new form of extreme summarization, for scientific papers. TLDR generation involves high source compression and requires expert background knowledge and understanding of complex domain-specific language. To facilitate study on this task, we introduce SCITLDR, a new multi-target dataset of 5.4K TLDRs over 3.2K papers. SCITLDR contains both author-written and expert-derived TLDRs, where the latter are collected using a novel annotation protocol that produces high-quality summaries while minimizing annotation burden. We propose CATTS, a simple yet effective learning strategy for generating TLDRs that exploits titles as an auxiliary training signal. CATTS improves upon strong baselines under both automated metrics and human evaluations. Data and code are publicly available at https://github.com/allenai/scitldr. 1 TLDR is an acronym that stands for "too long; didn't read," which is often used in online informal discussion (e.g., Twitter or Reddit) about scientific papers. For visual clarity, we omit the semi-colon.

For agents

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

get_harvested_code_for_paper("2004.15011")
get_code_for_paper("2004.15011")
have("2004.15011")

Connect an agent — have() is free.