Working out what this map is made of…
Working out what this map is made of…
This page says what the “embedding” map is made of, counted from the map itself at the moment you loaded it. Two kinds of claim sit in a map: quotes, which are checked word for word against the text this system pulled out of each paper, and the shape (which papers were grouped together, which relationships exist and which way round they point, which paper matters most), which is this tool’s reading and not the history of the term. The figures below say how much of each this map holds; none of them is a score, and there is no grade, no colour and no pass mark anywhere on the page.
You are reading version 1 of this map, which is the current one. It is the only published version so far. If the term is mapped again, this version keeps this address and the new one becomes current.
Beside each figure is the same figure for every other map in this system: one map per term, the current version of each, never two versions of the same term. With only a few maps to compare against, read those as points of comparison, not as an average or as what a map should show. n=4
How many papers this map has read, and how many of those define the term. These figures count only the papers that were actually read. A paper the run tried and failed to read is not counted anywhere on this page, so the number of papers the run set out to read may be higher than any figure here.
Papers whose text this map has read, whether or not a definition was found in them. A paper whose text could not be downloaded or read is not counted here at all.
What would change this: The map being rebuilt over more papers.
the other maps in this system: instrumental variable 145 · dark energy 145 · attention 156
Of those, the papers that give the term a definition of their own. Every percentage about definitions below is a share of this number.
What would change this: The map being rebuilt, and a paper being read differently the second time, in either direction.
the other maps in this system: instrumental variable 96 · dark energy 100 · attention 88
Papers read and found not to define the term. That is a finding, not a gap, and it is reported here rather than folded away. Shown so the percentages below can be read honestly: they are percentages of the papers that define the term, not of every paper read.
What would change this: The map being rebuilt.
the other maps in this system: instrumental variable 49 · dark energy 45 · attention 68
Whoever asks for a map may name up to three papers it should not miss. A named paper is added to what gets read and is then treated exactly like every other paper: it is read, its quotes are checked the same way, and a relationship to it is drawn only where one paper really does cite the other. Being named buys a reading and nothing else, so a named paper can end up defining nothing and connected to nothing, and this block says when one has.
What would change this: Only mapping the term again, and somebody naming different papers when they do.
Nobody named a paper for this map: every paper in it was chosen by this system. 0 of the 4 maps in this system have a paper somebody named.
How much of this map's evidence is currently matched against the text this system pulled out of each paper. Every quote is in exactly one of three states. Verified: found word for word in the text this system pulled out of the paper. Inferred: the check ran and could not find it. Not re-checked: the check has not yet run against the paper's current text. The three always add up to the total.
Definition quotes found word for word in the text this system pulled out of that paper, as that text stands now, with the share of the papers that define the term beside it. The share reads absent rather than 0% when no paper defines the term, because a share of nothing is not zero.
What would change this: The map being rebuilt. Also the text of a paper changing: any quote checked against the old text goes back to not re-checked.
the other maps in this system: instrumental variable 94 (97.9%) · dark energy 98 (98.0%) · attention 88 (100.0%)
Definition quotes where the check ran against the current text and did not find the passage.
What would change this: The map being rebuilt and the passage being found this time. Or the system reading the paper's text differently and losing it.
How the map's meanings and relationships are distributed. All of this is the tool's reading rather than anything checked against a paper: a grouping is a judgement, and so is the direction of a relationship.
Groups of papers this map treats as holding one meaning of the term.
What would change this: This changes if the map is rebuilt, or if someone editing the map splits, merges or deletes a group.
the other maps in this system: instrumental variable 54 · dark energy 7 · attention 32
Meanings with a single paper in them, with the share of every meaning in the map beside it. The share reads absent rather than 0% on a map with no meanings.
What would change this: Rebuilding the map, or someone editing it merging two meanings into one.
the other maps in this system: instrumental variable 32 (59.3%) · dark energy 3 (42.9%) · attention 19 (59.4%)
Each paper is dated by the earliest date we can trust for it: whichever is earlier of the date the catalogue gives and the date built into its arXiv number. Read it as when the work first appeared, not as a publication date. A paper with no year is counted separately and never guessed at. Two things are counted here and they are not the same: every paper this map read, and the ones that define the term.
Which version of this map you are reading, and what building that version took. The figures below come from the record kept while a map is being built for someone who asked for it, so a map built directly by the team, rather than through a request on this site, has none of them and says so. Every one of them describes the build behind this version and no other. Where a figure reads not recorded, nothing in this system holds it; it is not zero and it is not an estimate.
Mapping a term again does not replace the map that was there: it adds a version, and every version keeps its own address and stays readable, so a reader who cited one can still open the thing they read. One version at a time is the current one for a term, and that is the one the list of maps offers. Changes made by hand stay on the version they were made against and are never carried into a new one: a new build groups the papers afresh, so an edit moved across could end up attached to a group its author never saw.
What would change this: Only mapping the term again, which adds a version and makes whichever one was current the version before it. Nothing about this version changes when that happens except that it stops being the current one.
Which build of this term produced the map you are reading, and how many builds of it are published. Every figure on this page is counted from this version alone. The count is of versions a reader can open: a build that started and never finished takes a version number and publishes nothing, so the number this version was given can be higher than the number of versions there are.
Only mapping the term again.
Three things a reader would reasonably want that nothing in this system records. They are named here rather than left as blanks, because a blank reads as a zero.
Every figure here was counted from the map when you loaded this page. None of them is stored, so an edit to the map changes them at once, and nothing here can be out of date.
the other maps in this system: instrumental variable 2 · dark energy 2 · attention 0
Quotes whose last check ran against an older copy of the paper's text, or that have no record of a check at all. They are not claimed as verified until the check runs again.
What would change this: The check being run again, which costs nothing, because it only compares text with text.
the other maps in this system: instrumental variable 0 · dark energy 0 · attention 0
Relationships between two of this map's meanings, each carrying a quote from both papers.
What would change this: The map being rebuilt, or someone editing the map deleting a relationship.
the other maps in this system: instrumental variable 23 · dark energy 223 · attention 287
Relationships where both quotes were matched word for word against their papers' current text, with the share of every relationship in the map beside it. Both sides, not one: half a checked relationship is not a checked relationship. It counts relationships, not quotes: many relationships rest on the same quote. The share reads absent rather than 0% on a map with no relationships.
What would change this: The same check as above.
the other maps in this system: instrumental variable 19 (82.6%) · dark energy 212 (95.1%) · attention 286 (99.7%)
Relationships where the check ran against both papers' current text and at least one quote was not found.
What would change this: The check running again and finding both quotes this time.
the other maps in this system: instrumental variable 4 · dark energy 11 · attention 1
Relationships whose last check ran against an older copy of a paper's text, or that have no record of a check at all. A relationship whose paper is missing from our records sits here too, because a missing paper cannot support a check. On an older map this can read as the whole map, which is honest rather than broken.
What would change this: The check being run again.
the other maps in this system: instrumental variable 0 · dark energy 0 · attention 0
Every paper in this map came from the same place, “arxiv”. When a second source is added, these figures will be split by source here.
These 320 relationships rest on 125 different quoted passages, because each paper shows the same key quote wherever it appears. One passage is behind 35 of them, so the number of relationships is not a count of separate pieces of evidence.
Meanings that no relationship connects to a different meaning. These are what the evidence page lists as unconnected.
What would change this: A new relationship reaching one of them, or the map being rebuilt.
the other maps in this system: instrumental variable 30 · dark energy 0 · attention 0
The biggest single group (n4), by how many papers it holds.
What would change this: Rebuilding the map, or someone editing it splitting this group in two.
the other maps in this system: instrumental variable 7 · dark energy 49 · attention 19
Relationships with the largest meaning at either end, with the share of every relationship in the map beside it. It says how much of the map hangs off one group. The share reads absent rather than 0% on a map with no relationships.
What would change this: The same things as above, plus anything that adds or removes a relationship.
the other maps in this system: instrumental variable 2 (8.7%) · dark energy 187 (83.9%) · attention 115 (40.1%)
Where two groups of papers are linked more than once, the links do not always say the same thing: one paper may extend an idea while another narrows it. These are the relationships to read first if you are checking the tool's reading. The share reads absent rather than 0% on a map with no relationships.
What would change this: Someone editing the map changing what kind of link a relationship is, or the map being rebuilt.
the other maps in this system: instrumental variable 2 of 23 (8.7%), across 1 pairs · dark energy 192 of 223 (86.1%), across 8 pairs · attention 167 of 287 (58.2%), across 30 pairs
Every one of this map’s 320 typed relationships, counted by which kind of link it was given. A relationship carries exactly one kind, so these add up to that total. A kind no relationship was given reads 0 rather than being left out: having none of a kind is a finding, and a missing line is not.
What would change this: The map being rebuilt, or someone editing the map changing what kind a relationship is, or deleting one.
| kind | what it means | this map | instrumental variable | dark energy | attention |
|---|---|---|---|---|---|
| extends | builds on / generalises | 102 (31.9%) | 7 (30.4%) | 139 (62.3%) | 115 (40.1%) |
| contests | disputes | 4 (1.3%) | 2 (8.7%) | 9 (4.0%) | 18 (6.3%) |
| narrows | restricts to a special case | 60 (18.8%) | 8 (34.8%) | 73 (32.7%) | 42 (14.6%) |
| renames | same idea, new name | 15 (4.7%) | 0 (0.0%) | 2 (0.9%) | 3 (1.0%) |
| merges | fuses two ideas | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) |
| migrates | carries into a new field | 139 (43.4%) | 6 (26.1%) | 0 (0.0%) | 109 (38.0%) |
The oldest paper in the map that defines the term. The chart above can start further left, where a paper was read and did not define the term.
What would change this: The map being rebuilt and reaching further back, or a wrong year being corrected.
the other maps in this system: instrumental variable 2005 · dark energy 1988 · attention 2012
The newest paper in the map that defines the term. Read it beside the newest paper read, below: if that one is later, this map read newer work and did not find the term defined in it.
What would change this: Only mapping the term again brings in newer work.
the other maps in this system: instrumental variable 2025 · dark energy 2016 · attention 2021
The newest paper in the map, whether or not it defines the term, which is the right end of the chart above. It is never earlier than the latest year above, because the papers that define the term are some of the papers read.
What would change this: The map being rebuilt over papers chosen differently.
the other maps in this system: instrumental variable 2025 · dark energy 2016 · attention 2021
How many of this map's papers were written in the last five years, which is 2022 to 2026 as this page was loaded. This is what the map chose to read, before anything was found in them. It is a count of what was selected and not a result: changing how papers are chosen moves it by construction, so it can say what this map read and it cannot say whether the reading was any good.
What would change this: The map being rebuilt, and the way papers are chosen for it changing. The window moves too: it is the five calendar years ending in the year you loaded this page.
the other maps in this system: instrumental variable 32 · dark energy 0 · attention 0
Of the papers above, the ones found to give the term a definition of their own. A paper read and found not to define the term is a finding about the literature, not a gap in the map. Unlike the figure above, this one can come back lower after more recent papers are read, because whether a paper defines the term is decided by reading it.
What would change this: The map being rebuilt, and a paper being read differently the second time, in either direction.
the other maps in this system: instrumental variable 23 of 32 · dark energy 0 · attention 0
Papers that define the term and have no year at all. They are nowhere in the chart above and in none of the figures here, rather than being given a guessed date.
What would change this: Only better publication dates for those papers. Nothing about the map itself moves this.
the other maps in this system: instrumental variable 0 · dark energy 0 · attention 0
A map can be built in two passes: the first reads papers chosen before anything is known about the term, and the second asks the map it has just built what came after its turning points, then reads those too. Both happen before the map is shown to anyone. What follows says what the second pass read and what was found in it, separately from the first, so that what the second pass bought can be told apart from what the first would have found anyway.
What would change this: Only mapping the term again. A map is built in one pass or two and is never added to afterwards.
This map was built in one pass: nothing in its record marks a paper as read in a second one. 0 of the 4 maps in this system were built in more than one pass, so a second pass’s reading is not yet something this page can show you.
the other maps in this system: instrumental variable version 1, the current one · dark energy version 1, the current one · attention version 1, the current one
The day this version stopped being the current map for its term. It is the date of the change and not a judgement about the version: a replaced version is what this tool read at that time, it is not withdrawn, and every quote in it is still checked the same way. Where it says nothing records it being replaced, this version is not the current one and nothing in this system says when that changed or what replaced it.
What would change this: Only mapping the term again, which replaces whichever version is current at that moment.
the other maps in this system: instrumental variable not replaced: this is the current version · dark energy not replaced: this is the current version · attention not replaced: this is the current version
Renames, splits, merges and deletions somebody made to this version after it was built. It counts this version only and is never added up across versions, because two versions group the papers differently and a total across them would be a number about nothing. A version showing none may still have a sibling that carries edits, and the table below says which.
What would change this: Somebody editing this version. A change made to another version never appears here.
the other maps in this system: instrumental variable 0 · dark energy 0 · attention 0
| version | replaced on | changes made by hand | ||
|---|---|---|---|---|
| 1 | the one you are reading, current | not replaced | none | this page |
Completed runs this system has a record of for this map. The figures below describe the most recent one, not the map's whole life. A map the team built directly, rather than through a request on this site, has no run record, and every figure below then reads not recorded.
What would change this: Mapping the term again.
the other maps in this system: instrumental variable 1 · dark energy 1 · attention 1
From the run starting to the run finishing, for the most recent completed run.
What would change this: How many papers the run had to read, and how many of those this system had already read for an earlier map.
the other maps in this system: instrumental variable 46 min 18 s · dark energy 52 min 59 s · attention 18 min 16 s
From the map being asked for to the run finishing, which includes waiting in the queue.
What would change this: Everything above, plus how long the request waited before it started.
the other maps in this system: instrumental variable 99 min 2 s · dark energy 53 min 3 s · attention 611 min 22 s