{"id":1418,"date":"2026-09-17T17:10:53","date_gmt":"2026-09-17T08:10:53","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/glossary-benchmarks-en\/"},"modified":"2026-09-18T11:57:22","modified_gmt":"2026-09-18T02:57:22","slug":"glossary-benchmarks-en","status":"publish","type":"page","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-benchmarks-en\/","title":{"rendered":"LLM Benchmark Glossary: How to Read the Scores"},"content":{"rendered":"

What the benchmark scores quoted in our articles actually measure, how to read the number, and whether higher or lower is better. This is the same glossary our pipeline hands to the article writer: a benchmark that is not listed here is never explained in an article, only compared. Definitions are maintained by hand in the site’s source code, not generated by an AI model.<\/p>\n

Two things to keep in mind. First, error rates (WER, CER) and perplexity are lower-is-better<\/strong> \u2014 reading them like accuracy flips the conclusion. Second, scores from different model cards are usually not directly comparable: prompt format, number of samples and judging method differ between reports.<\/p>\n

Coding<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
HumanEval<\/td>\nHumanEval+, EvalPlus<\/td>\nWhether the model can implement short Python functions from a docstring specification; graded automatically with unit tests.<\/td>\nPass@1 = share of problems solved on the first attempt (%)<\/td>\nHigher<\/td>\n<\/tr>\n
MBPP<\/td>\nMBPP+<\/td>\nWhether the model can solve entry- to intermediate-level Python tasks.<\/td>\nPass@1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n
BigCodeBench<\/td>\n\u2014<\/td>\nPractical Python tasks that require calling several libraries correctly.<\/td>\nPass@1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n
LiveCodeBench<\/td>\n\u2014<\/td>\nFresh competitive-programming problems collected after a cutoff date, designed to avoid training-data contamination.<\/td>\nPass@1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Coding (real repositories)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
SWE-bench<\/td>\nSWE-bench Verified, SWE-bench Pro, SWE-bench Multimodal<\/td>\nWhether the model can fix real GitHub issues by editing a repository; a hidden test suite decides.<\/td>\nResolved = share of issues fixed (%)<\/td>\nHigher<\/td>\n<\/tr>\n
Aider Polyglot<\/td>\nAider<\/td>\nWhether the model can fix code in several languages by editing existing files.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Coding (competitive)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
Codeforces<\/td>\nCodeForces<\/td>\nCompetitive-programming skill.<\/td>\nRating (Elo) on the same scale as human contestants \u2014 not a percentage<\/strong><\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Agents (terminal)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
Terminal-Bench<\/td>\nTerminalBench, Terminal Bench<\/td>\nWhether the model can complete a task end-to-end by running commands in a real terminal.<\/td>\nPass@1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Agents (tool use)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
\u03c4-bench<\/td>\ntau-bench, TAU-bench, tau2-bench<\/td>\nWhether the model can follow business policies while talking to a user and calling tools.<\/td>\nSuccess rate (%)<\/td>\nHigher<\/td>\n<\/tr>\n
BFCL<\/td>\nBerkeley Function Calling Leaderboard, Berkeley Function-Calling Leaderboard<\/td>\nHow accurately the model chooses functions (tools) and builds their arguments.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Agents (research)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
GAIA<\/td>\n\u2014<\/td>\nReal-world research questions that require tools such as search and file handling.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Agents (GUI)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
OSWorld<\/td>\n\u2014<\/td>\nWhether the model can complete tasks by operating a real operating-system desktop.<\/td>\nSuccess rate (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Math<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
GSM8K<\/td>\n\u2014<\/td>\nGrade-school word problems solved step by step; close to saturated for top models, so differences are small.<\/td>\nEM = exact match of the final answer (%)<\/td>\nHigher<\/td>\n<\/tr>\n
MATH<\/td>\nMATH-500, MATH500<\/td>\nHigh-school to competition-level mathematics.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Math (competition)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
AIME<\/td>\nAIME 2024, AIME 2025<\/td>\nProblems from the American Invitational Mathematics Examination, a US olympiad qualifier.<\/td>\nPass@1 (%). Each exam has only 15 problems, so one problem moves the score a lot<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Multilingual<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
MGSM<\/td>\n\u2014<\/td>\nGSM8K problems translated into many languages, to see whether math ability holds outside English.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

General knowledge<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
MMLU<\/td>\n\u2014<\/td>\nMultiple-choice questions across 57 subjects; the classic measure of broad, shallow knowledge. Top models are close together.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

General knowledge and reasoning<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
MMLU-Pro<\/td>\nMMLU Pro<\/td>\nA harder MMLU with ten answer choices and more reasoning-heavy questions.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n
AGIEval<\/td>\n\u2014<\/td>\nExams written for humans, such as university entrance and professional qualification tests.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Hard science questions<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
GPQA Diamond<\/td>\nGPQA-Diamond, GPQA<\/td>\nPhD-level physics, chemistry and biology questions written to resist web search (Diamond is the hardest subset).<\/td>\nPass@1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Extremely hard questions<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
HLE<\/td>\nHumanity’s Last Exam, Humanitys Last Exam<\/td>\nExpert-written questions across many fields; every model still scores low, so a few dozen percent is already high.<\/td>\nPass@1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Reasoning<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
BBH<\/td>\nBIG-Bench Hard<\/td>\nThe BIG-Bench tasks on which models of the time trailed humans.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n
BBEH<\/td>\nBIG-Bench Extra Hard<\/td>\nA harder successor to BBH, created after BBH saturated.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Expert knowledge<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
SuperGPQA<\/td>\n\u2014<\/td>\nGraduate-level questions across a much wider range of disciplines.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Multilingual (Chinese)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
C-Eval<\/td>\nCEval, CMMLU<\/td>\nMultiple-choice academic and professional questions in Chinese.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Reading comprehension<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
DROP<\/td>\n\u2014<\/td>\nReading questions that require picking numbers out of a passage and adding, subtracting or comparing them.<\/td>\nF1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Commonsense reasoning<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
HellaSwag<\/td>\n\u2014<\/td>\nChoosing the natural continuation of a sentence. An old benchmark, saturated for top models.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n
Winogrande<\/td>\nWinoGrande<\/td>\nResolving which noun a pronoun refers to from context.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n
ARC-Challenge<\/td>\nARC-c, AI2 Reasoning Challenge<\/td>\nGrade-school science questions that are hard to answer by retrieval alone.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Factuality<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
SimpleQA<\/td>\nSimpleQA-Verified<\/td>\nShort factual questions with a single correct answer; tests whether the model makes things up instead of admitting ignorance.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n
TruthfulQA<\/td>\n\u2014<\/td>\nWhether the model avoids repeating common misconceptions.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Instruction following<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
IFEval<\/td>\n\u2014<\/td>\nWhether formal instructions such as \u201cunder 300 words\u201d or \u201canswer in JSON\u201d are followed, independent of content.<\/td>\nCompliance rate (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Chat quality<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
Arena-Hard<\/td>\nArenaHard, MT-Bench, MTBench<\/td>\nResponses to difficult chat prompts, judged relatively by another LLM.<\/td>\nWin rate or score. The judge is also an LLM, so absolute values are not comparable across reports<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Long context<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
LongBench<\/td>\nLongBench-v2, LongBench-V2<\/td>\nSummarising, answering questions about and understanding code in long documents.<\/td>\nEM\/F1 (%)<\/td>\nHigher<\/td>\n<\/tr>\n
RULER<\/td>\n\u2014<\/td>\nHow much of the advertised context length is actually usable.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n
Needle in a Haystack<\/td>\nNIAH, Needle-in-a-Haystack<\/td>\nWhether one sentence hidden somewhere in a long context can be retrieved.<\/td>\nRetrieval rate (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Image understanding<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
MMMU<\/td>\nMMMU-Pro<\/td>\nCollege-level questions that include charts and photos, answered from the image.<\/td>\nEM (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Image understanding (documents)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
DocVQA<\/td>\n\u2014<\/td>\nAnswering questions about document images such as invoices and forms.<\/td>\nANLS \/ accuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Image understanding (charts)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
ChartQA<\/td>\n\u2014<\/td>\nReading values and trends from chart images.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Image understanding (math)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
MathVista<\/td>\n\u2014<\/td>\nMath problems that involve figures and graphs, solved from the image.<\/td>\nAccuracy (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Image understanding (grounding)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
RefCOCO<\/td>\nRefCOCO+, RefCOCOg<\/td>\nLocating an object in an image from a description such as \u201cthe second red car from the left\u201d.<\/td>\nAcc@0.5 = correct if overlap (IoU) with the reference box is at least 0.5 (%)<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Image understanding (text)<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
OCRBench<\/td>\n\u2014<\/td>\nReading text embedded in images.<\/td>\nScore<\/td>\nHigher<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Speech recognition<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
WER<\/td>\nWord Error Rate<\/td>\nWord-level error rate of transcriptions.<\/td>\nLower is better<\/strong> (%)<\/td>\nLower<\/td>\n<\/tr>\n
CER<\/td>\nCharacter Error Rate<\/td>\nCharacter-level error rate of transcriptions.<\/td>\nLower is better<\/strong> (%)<\/td>\nLower<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Raw language-model accuracy<\/h2>\n
\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n
Benchmark<\/th>\nAlso written as<\/th>\nWhat it measures<\/th>\nHow to read the number<\/th>\nBetter<\/th>\n<\/tr>\n<\/thead>\n
Perplexity<\/td>\nPPL<\/td>\nHow well the model predicts the next token; commonly used to compare quality loss from quantization.<\/td>\nLower is better<\/strong><\/td>\nLower<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

45 benchmarks. Last updated 2026-09-18 (JST). See also our quantization and model-format glossary<\/a>.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

What the benchmark scores quoted in our articles actually measure, how to read the number, and whether higher or lower is better. This is the same glossary our pipeline hands to the article writer: a benchmark that is not listed here is never explained in an article, only compared. Definitions are maintained by hand in […]<\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-1418","page","type-page","status-publish","hentry"],"lang":"en","translations":{"en":1418,"ja":1416},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/pages\/1418","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=1418"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/pages\/1418\/revisions"}],"predecessor-version":[{"id":1558,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/pages\/1418\/revisions\/1558"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=1418"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}