1
Your word
measured
“a soldier must be prepared to kill his enemies”
Definitions come from WordNet, a hand-curated dictionary.
an armed adversary (especially a member of an opposing military force)
仇敌 is Chinese for foe, and every measurement below was made on that English word. Definitions come from English WordNet; Open Multilingual WordNet supplies the Chinese headword, not a Chinese definition.
Panels 3, 4 and 5 are computed per model. Switch to see them disagree.
“a soldier must be prepared to kill his enemies”
Definitions come from WordNet, a hand-curated dictionary.
6 of those are a raw byte rather than a character: Mistral 7B splits this word below the character, so no single token it sees is readable.
This panel splits 仇敌 itself, not foe — a tokenizer does not care what language it is fed. It is the only panel here that does.
This is the tokenizer splitting text, not the model understanding it.
friend is the dictionary opposite of foe, yet this model puts it closer than 1 of the 2 synonyms it can score. Opposites share the sentences a word lives in, so cosine alone cannot tell them apart - which is why the dictionary seeds this list rather than the geometry.
Shade shows how close the model puts each word to foe. 2 of 3 dictionary synonyms are in this build; the rest have no vector to compare yet.
Shade shows how close the model puts each word to foe, 1 of 8 also appear in the dictionary. 2 of 3 dictionary synonyms are in this build; the rest have no vector to compare yet.
This model splits foe into 2 pieces, so it has no vector of its own here: this is the average of its fragments, and the results below are correspondingly rough.
friend is the dictionary opposite of foe, yet this model puts it closer than 1 of the 2 synonyms it can score. Opposites share the sentences a word lives in, so cosine alone cannot tell them apart - which is why the dictionary seeds this list rather than the geometry.
Shade shows how close the model puts each word to foe. 2 of 3 dictionary synonyms are in this build; the rest have no vector to compare yet.
Model neighbours are distributional, not dictionary synonyms. Two words can be close because they appear in similar sentences, which is why an antonym can outscore a synonym here.
This model splits foe into 2 pieces, so it has no vector of its own here: this is the average of its fragments, and the results below are correspondingly rough.
Scores are projections onto axes we defined from anchor words, not labels the model assigns.
These sit just outside the closest neighbours in panel 3, and no dictionary lists any of them as related to foe. That gap is the model’s own learned association.
These sit just outside the closest neighbours in panel 3, and no dictionary lists any of them as related to foe. That gap is the model’s own learned association.
This model splits foe into 2 pieces, so it has no vector of its own here: this is the average of its fragments, and the results below are correspondingly rough.
These sit just outside the closest neighbours in panel 3, and no dictionary lists any of them as related to foe. That gap is the model’s own learned association.
These are statistical associations in the training data, not the model thinking.