SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.07540 · EMNLP · 2022

SCIENCEWORLD: Is your Agent Smarter than a 5 th Grader?

Prithviraj Ammanabrolu, Marc-Alexandre Côté, Peter Jansen, Ruoyao Wang

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 7 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/macaw canonical 6 of 7
allenai/scienceworld canonical 1 of 1
FunctionStatusWhere it lives
build_simplification_str Ran allenai/scienceworld/examples/random_agent.py
code served (permissive licence) · get_code("3c4d106b37c63938")
fix_t5_unk_characters Ran allenai/macaw/macaw/eval_utils.py
code served (permissive licence) · get_code("20633ae7459cfada")
load_jsonl Ran allenai/macaw/macaw/utils.py
code served (permissive licence) · get_code("c30d329a797e15bd")
replace_punctuation Ran allenai/macaw/macaw/eval_utils.py
code served (permissive licence) · get_code("d2166ef001a9e416")
run_model Ran allenai/macaw/macaw/utils.py
code served (permissive licence) · get_code("0ff0cabbc0770012")
run_model_with_outputs Ran allenai/macaw/macaw/batch_eval.py
code served (permissive licence) · get_code("df88f11724e3cab5")
score_string_similarity Ran allenai/macaw/macaw/eval_utils.py
code served (permissive licence) · get_code("00139853f0c3a89e")
load_model Not yet run allenai/macaw/macaw/utils.py
code served (permissive licence) · get_code("a96d409802f5ecdc")

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 present SCIENCEWORLD, a benchmark to test agents' scientific reasoning abilities in a new interactive text environment at the level of a standard elementary school science curriculum. Despite the transformer-based progress seen in question-answering and scientific text processing, we find that current models cannot reason about or explain learned science concepts in novel contexts. For instance, models can easily answer what the conductivity of a known material is but struggle when asked how they would conduct an experiment in a grounded environment to find the conductivity of an unknown material. This begs the question of whether current models are simply retrieving answers by way of seeing a large number of similar examples or if they have learned to reason about concepts in a reusable manner. We hypothesize that agents need to be grounded in interactive environments to achieve such reasoning capabilities. Our experiments provide empirical evidence supporting this hypothesisshowing that a 1.5 million parameter agent trained interactively for 100k steps outperforms a 11 billion parameter model statically trained for scientific question-answering and reasoning from millions of expert demonstrations.

For agents

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

get_harvested_code_for_paper("2203.07540")
get_code_for_paper("2203.07540")
have("2203.07540")

Connect an agent — have() is free.