Papers › Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering

Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering

20 Sep 2022arXiv:2209.09513archive 2025-07-28

Pan Lu, Swaroop Mishra, Tony Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, Ashwin Kalyan

When answering a question, humans utilize the information available across different modalities to synthesize a consistent and complete chain of thought (CoT). This process is normally a black box in the case of deep learning models like large-scale language models. Recently, science question benchmarks have been used to diagnose the multi-hop reasoning ability and interpretability of an AI system. However, existing datasets fail to provide annotations for the answers, or are restricted to the textual-only modality, small scales, and limited domain diversity. To this end, we present Science Question Answering (ScienceQA), a new benchmark that consists of ~21k multimodal multiple choice questions with a diverse set of science topics and annotations of their answers with corresponding lectures and explanations. We further design language models to learn to generate lectures and explanations as the chain of thought (CoT) to mimic the multi-hop reasoning process when answering ScienceQA questions. ScienceQA demonstrates the utility of CoT in language models, as CoT improves the question answering performance by 1.20% in few-shot GPT-3 and 3.99% in fine-tuned UnifiedQA. We also explore the upper bound for models to leverage explanations by feeding those in the input; we observe that it improves the few-shot performance of GPT-3 by 18.96%. Our analysis further shows that language models, similar to humans, benefit from explanations to learn from fewer data and achieve the same performance with just 40% of the data. The data and code are available at https://scienceqa.github.io.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 5 code samples: 3 from 1 repository linked to this paper, 2 identical to code harvested from another paper's repository. Across all 5: 4 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 4 where Syntology's instrument failed) · 1 unverified. Where Syntology's instrument failed: 4 where our contract draft was wrong; failures of Syntology's instrument, not of the code. Every run on this page is a failure of Syntology's instrument, not of the code.

By repository: official repository: 3 samples from 1 repository, 2 ran (0 with no instrument failure; 2 where Syntology's instrument failed); 2 identical to code first harvested elsewhere. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

lupantech/ScienceQA officialmentioned on GitHubpytorchSyntology: 3 samples, 2 ran (2 instrument-failed) · 1 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

5 samples harvested: 4 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 4 where Syntology's instrument failed) · 1 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Every run on this page is a failure of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

4ran · our draft was wrong
1unverified

Licence: 2 of the 5 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from lupantech/ScienceQA. 2 samples are listed last, in a group without a repository. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

lupantech/ScienceQA official repository 3 samples, 2 ran (2 instrument-failed) · 1 unverified
get_acc_with_contion tools/evaluate_acc.py ran · our draft was wrong MIT (permissive) · bef96bae3f74e855 · report fixture: signature res_pd:DataFrame/0/?, key:str/0/str, values:list/1/str · returned str · level recorded 2026-09-22
get_question_text in 5 papers models/base_prompt.py ran · our draft was wrong MIT (permissive) · c86a79e229253816 · report fixture: signature problem:dict/0/? · returned str · level recorded 2026-09-22
get_scores in 22 papers tools/evaluate_acc.py unverified MIT (permissive) · ab3717114ec507c3 · report
Repository not located 2 samples, 2 ran (2 instrument-failed) · 0 unverified
2 samples are identical to code Syntology harvested from another paper's repository; this paper's own copy was not located, so no repository is named and no licence is recorded.
get_choice_text in 7 papers repository not located ran · our draft was wrong licence of this copy not recorded · 7a8dffc83d89cedd · report fixture: signature probelm:dict/0/?, options:list/1/str · returned str · level recorded 2026-09-22
get_context_text in 7 papers repository not located ran · our draft was wrong licence of this copy not recorded · 85f084f6054a421e · report fixture: signature problem:dict/0/?, use_caption:bool/0/bool · returned str · level recorded 2026-09-22

Tasks

Multimodal Deep LearningMultimodal ReasoningMultiple-choiceOpen-Domain Question AnsweringQuestion AnsweringScience Question AnsweringVisual Commonsense ReasoningVisual Question Answering (VQA)

Datasets

Introduced by this paper, per the archive.

ScienceQA

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Avg. Accuracy 75.17 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Grades 1-6 78.23 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Grades 7-12 69.68 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Image Context 67.43 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Language Science 78.09 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Natural Science 75.44 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) No Context 79.93 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Social Science 70.87 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT (QCM→ALE , 2-shot) Text Context 74.68 #6 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Avg. Accuracy 74.61 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Grades 1-6 78.49 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Grades 7-12 67.63 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Image Context 66.09 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Language Science 77.55 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Natural Science 76.60 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) No Context 79.58 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Social Science 65.92 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 - CoT(QCM→AE, 2-shot) Text Context 75.51 #7 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Avg. Accuracy 74.11 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Grades 1-6 77.06 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Grades 7-12 68.82 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Image Context 66.53 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Language Science 78.91 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Natural Science 71.00 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) No Context 81.81 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Social Science 76.04 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA UnifiedQA-BASE - CoT (QCM→ALE) Text Context 66.42 #8 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Avg. Accuracy 73.97 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Grades 1-6 76.80 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Grades 7-12 68.89 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Image Context 67.28 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Language Science 76.00 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Natural Science 74.64 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) No Context 77.42 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Social Science 69.74 #9 of 10 Archive leaderboard report
Science Question Answering ScienceQA GPT-3 (QCM→A, 2-shot) Text Context 74.44 #9 of 10 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdamCosine AnnealingDropoutGPT-3Layer NormalizationLinear Warmup With Cosine AnnealingMulti-Head AttentionResidual ConnectionSoftmax

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections