Dataset Viewer
Auto-converted to Parquet Duplicate
instruction
stringlengths
16
91
answer
stringclasses
36 values
tag
stringclasses
34 values
split_kind
stringclasses
8 values
text
stringlengths
53
134
What is the assistant's job?
I answer the user's request directly.
assistant_role
train_seen
Instruction: What is the assistant's job? Answer: I answer the user's request directly. <END>
Context: name = Riley; password = orange. Question: What is the password?
orange.
context_password
context_seen_train
Instruction: Context: name = Riley; password = orange. Question: What is the password? Answer: orange. <END>
What favorite film did I tell you?
I don't know from the given information.
idk_favorite_movie
task_upstream_regression_train
Instruction: What favorite film did I tell you? Answer: I don't know from the given information. <END>
The name supplied in this prompt is Jordan. Question: Which name was supplied?
Jordan.
context_name
context_seen_train
Instruction: The name supplied in this prompt is Jordan. Question: Which name was supplied? Answer: Jordan. <END>
User: Call me Jordan. User: What should you call me?
Jordan.
context_name
context_seen_train
Instruction: User: Call me Jordan. User: What should you call me? Answer: Jordan. <END>
Give the code/prose label: return user.email
code
classify_code
task_upstream_regression_train
Instruction: Give the code/prose label: return user.email Answer: code <END>
Is this code or prose: This sentence is plain English.
prose
classify_prose
task_seen_train
Instruction: Is this code or prose: This sentence is plain English. Answer: prose <END>
Give the sentiment label only: The fix worked.
positive
sentiment_positive
task_seen_train
Instruction: Give the sentiment label only: The fix worked. Answer: positive <END>
Classify this as positive or negative: This is awful.
negative
anti_collision_sentiment_negative_not_positive
task_upstream_regression_train
Instruction: Classify this as positive or negative: This is awful. Answer: negative <END>
Classify this as positive or negative: I really like this.
positive
sentiment_positive
task_upstream_regression_train
Instruction: Classify this as positive or negative: I really like this. Answer: positive <END>
Context: name = Evan; password = purple. Question: What is the password?
purple.
context_password
context_seen_train
Instruction: Context: name = Evan; password = purple. Question: What is the password? Answer: purple. <END>
Context: password = black. Question: What is the password?
black.
context_password
context_seen_train
Instruction: Context: password = black. Question: What is the password? Answer: black. <END>
Sentiment of this text: I love this tool.
positive
sentiment_positive
task_seen_train
Instruction: Sentiment of this text: I love this tool. Answer: positive <END>
Label the input type: x = add(1, 2)
code
classify_code
task_seen_train
Instruction: Label the input type: x = add(1, 2) Answer: code <END>
Label the input type: The schedule changed again.
prose
classify_prose
task_seen_train
Instruction: Label the input type: The schedule changed again. Answer: prose <END>
Tell me what Evan ate today.
I don't know from the given information.
idk_food_today
task_seen_train
Instruction: Tell me what Evan ate today. Answer: I don't know from the given information. <END>
Context: name: Kai. Question: Which name was given?
Kai.
context_name
context_seen_train
Instruction: Context: name: Kai. Question: Which name was given? Answer: Kai. <END>
Is this sentiment positive or negative: That was not useful.
negative
sentiment_negative
task_seen_train
Instruction: Is this sentiment positive or negative: That was not useful. Answer: negative <END>
What food did Evan eat today?
I don't know from the given information.
idk_food_today
task_seen_train
Instruction: What food did Evan eat today? Answer: I don't know from the given information. <END>
Write subtract(a, b).
def subtract(a, b): return a - b
code_subtract
train_seen
Instruction: Write subtract(a, b). Answer: def subtract(a, b): return a - b <END>
Decide if this is code or prose: user.name
code
classify_code
task_seen_train
Instruction: Decide if this is code or prose: user.name Answer: code <END>
What type of text is this: x = add(1, 2)
code
classify_code
task_upstream_regression_train
Instruction: What type of text is this: x = add(1, 2) Answer: code <END>
Remember this name: Jordan. Question: What name should you remember?
Jordan.
context_name
context_seen_train
Instruction: Remember this name: Jordan. Question: What name should you remember? Answer: Jordan. <END>
Give the code/prose label: The plan changed again.
prose
classify_prose
task_upstream_regression_train
Instruction: Give the code/prose label: The plan changed again. Answer: prose <END>
Answer only positive or negative: The result is excellent.
positive
sentiment_positive
task_seen_train
Instruction: Answer only positive or negative: The result is excellent. Answer: positive <END>
Classify sentiment: That was not useful.
negative
sentiment_negative
task_seen_train
Instruction: Classify sentiment: That was not useful. Answer: negative <END>
What is my favorite film?
I don't know from the given information.
idk_favorite_movie
task_seen_train
Instruction: What is my favorite film? Answer: I don't know from the given information. <END>
Classify this as positive or negative: This is a strong result.
positive
sentiment_positive
task_upstream_regression_train
Instruction: Classify this as positive or negative: This is a strong result. Answer: positive <END>
Classify this as positive or negative: That failed again.
negative
sentiment_negative
task_upstream_regression_train
Instruction: Classify this as positive or negative: That failed again. Answer: negative <END>
What is the usual color of a clear daytime sky?
Blue.
sky_color
train_seen
Instruction: What is the usual color of a clear daytime sky? Answer: Blue. <END>
Type label: for i in range(3): print(i)
code
classify_code
task_seen_train
Instruction: Type label: for i in range(3): print(i) Answer: code <END>
This prompt supplies the name Riley. Question: Which name is supplied?
Riley.
context_name
context_seen_train
Instruction: This prompt supplies the name Riley. Question: Which name is supplied? Answer: Riley. <END>
Choose the sentiment label for: That was useful.
positive
sentiment_positive
task_seen_train
Instruction: Choose the sentiment label for: That was useful. Answer: positive <END>
Line one says PASSWORD=gold. Question: What is PASSWORD?
gold.
context_password
context_upstream_regression_train
Instruction: Line one says PASSWORD=gold. Question: What is PASSWORD? Answer: gold. <END>
Classify this as positive or negative: I am happy with this.
positive
sentiment_positive
task_seen_train
Instruction: Classify this as positive or negative: I am happy with this. Answer: positive <END>
Classify input: user.name
code
classify_code
task_seen_train
Instruction: Classify input: user.name Answer: code <END>
Conversation: User: Use blue as the password. User: What password did I give?
blue.
context_password
context_seen_train
Instruction: Conversation: User: Use blue as the password. User: What password did I give? Answer: blue. <END>
Answer only code or prose: return value
code
classify_code
task_seen_train
Instruction: Answer only code or prose: return value Answer: code <END>
Answer only code or prose: y = multiply(3, 4)
code
classify_code
task_seen_train
Instruction: Answer only code or prose: y = multiply(3, 4) Answer: code <END>
Before changing code, what should happen first?
Inspect the current files first.
inspect_first
train_seen
Instruction: Before changing code, what should happen first? Answer: Inspect the current files first. <END>
Sentiment label only: This is strong.
positive
sentiment_positive
task_upstream_regression_train
Instruction: Sentiment label only: This is strong. Answer: positive <END>
Classify input: x = add(1, 2)
code
anti_collision_classify_code_not_prose
task_upstream_regression_train
Instruction: Classify input: x = add(1, 2) Answer: code <END>
Facts: password is orange. name is Riley. Question: Which name was given?
Riley.
context_name
context_seen_train
Instruction: Facts: password is orange. name is Riley. Question: Which name was given? Answer: Riley. <END>
Context: password: blue. Question: Which password was given?
blue.
context_password
context_seen_train
Instruction: Context: password: blue. Question: Which password was given? Answer: blue. <END>
Remember inside this prompt: password is blue. Question: What is the password?
blue.
context_password
context_seen_train
Instruction: Remember inside this prompt: password is blue. Question: What is the password? Answer: blue. <END>
Line one says NAME=Evan. Question: What is NAME?
Evan.
context_name
context_upstream_regression_train
Instruction: Line one says NAME=Evan. Question: What is NAME? Answer: Evan. <END>
Use this stored name value: Kai. Question: What stored name value should be used?
Kai.
context_name
context_upstream_regression_train
Instruction: Use this stored name value: Kai. Question: What stored name value should be used? Answer: Kai. <END>
Facts: name is Riley. password is orange. Question: Which password was given?
orange.
context_password
context_seen_train
Instruction: Facts: name is Riley. password is orange. Question: Which password was given? Answer: orange. <END>
Write the ADD function.
def add(a, b): return a + b
code_add
train_seen
Instruction: Write the ADD function. Answer: def add(a, b): return a + b <END>
Why do tests matter?
Tests show whether the code still behaves as expected.
tests_matter
train_seen
Instruction: Why do tests matter? Answer: Tests show whether the code still behaves as expected. <END>
Is this sentiment positive or negative: That was helpful.
positive
sentiment_positive
task_seen_train
Instruction: Is this sentiment positive or negative: That was helpful. Answer: positive <END>
Context: name = Jordan; password = green. Question: What is the name?
Jordan.
context_name
context_seen_train
Instruction: Context: name = Jordan; password = green. Question: What is the name? Answer: Jordan. <END>
Make this concise: I am going to arrive late today.
I will be late today.
anti_collision_rewrite_no_loop
task_upstream_regression_train
Instruction: Make this concise: I am going to arrive late today. Answer: I will be late today. <END>
Facts: password is purple. name is Alex. Question: Which name was given?
Alex.
context_name
context_seen_train
Instruction: Facts: password is purple. name is Alex. Question: Which name was given? Answer: Alex. <END>
Sentiment of this text: This is good.
positive
sentiment_positive
task_seen_train
Instruction: Sentiment of this text: This is good. Answer: positive <END>
What is my favorite movie?
I don't know from the given information.
idk_favorite_movie
task_seen_train
Instruction: What is my favorite movie? Answer: I don't know from the given information. <END>
Label the input type: y = multiply(3, 4)
code
classify_code
task_seen_train
Instruction: Label the input type: y = multiply(3, 4) Answer: code <END>
User: Use the name Nina. User: What name did I give?
Nina.
context_name
context_seen_train
Instruction: User: Use the name Nina. User: What name did I give? Answer: Nina. <END>
Answer only code or prose: user.name
code
classify_code
task_seen_train
Instruction: Answer only code or prose: user.name Answer: code <END>
Create add so it computes a plus b.
def add(a, b): return a + b
code_add
train_seen
Instruction: Create add so it computes a plus b. Answer: def add(a, b): return a + b <END>
Give the code/prose label: The schedule changed again.
prose
classify_prose
task_upstream_regression_train
Instruction: Give the code/prose label: The schedule changed again. Answer: prose <END>
Classify this as code or prose: total = add(x, y)
code
classify_code
task_seen_train
Instruction: Classify this as code or prose: total = add(x, y) Answer: code <END>
Classify input: z = divide(9, 3)
code
classify_code
task_seen_train
Instruction: Classify input: z = divide(9, 3) Answer: code <END>
Give the code/prose label: return user.email
code
classify_code
task_upstream_regression_train
Instruction: Give the code/prose label: return user.email Answer: code <END>
What sentiment is this: I really like this.
positive
sentiment_positive
task_upstream_regression_train
Instruction: What sentiment is this: I really like this. Answer: positive <END>
Make a Python function named subtract that subtracts b from a.
def subtract(a, b): return a - b
code_subtract
train_seen
Instruction: Make a Python function named subtract that subtracts b from a. Answer: def subtract(a, b): return a - b <END>
Positive or negative sentiment: That failed again.
negative
sentiment_negative
task_upstream_regression_train
Instruction: Positive or negative sentiment: That failed again. Answer: negative <END>
Classify input: The cat sat down.
prose
classify_prose
task_seen_train
Instruction: Classify input: The cat sat down. Answer: prose <END>
What sentiment is this: I hate this bug.
negative
sentiment_negative
task_seen_train
Instruction: What sentiment is this: I hate this bug. Answer: negative <END>
Give only the type label for this: return value
code
classify_code
task_seen_train
Instruction: Give only the type label for this: return value Answer: code <END>
Classify sentiment: This helped a lot.
positive
sentiment_positive
task_seen_train
Instruction: Classify sentiment: This helped a lot. Answer: positive <END>
Facts: name is Riley. password is silver. Question: Which password was given?
silver.
context_password
context_seen_train
Instruction: Facts: name is Riley. password is silver. Question: Which password was given? Answer: silver. <END>
Classify this as code or prose: The file is on the desk.
prose
classify_prose
task_seen_train
Instruction: Classify this as code or prose: The file is on the desk. Answer: prose <END>
Choose the sentiment label for: I am unhappy with this.
negative
sentiment_negative
task_seen_train
Instruction: Choose the sentiment label for: I am unhappy with this. Answer: negative <END>
Classify this as code or prose: The file is on the desk.
prose
classify_prose
task_seen_train
Instruction: Classify this as code or prose: The file is on the desk. Answer: prose <END>
Label the sentiment: I hate this bug.
negative
sentiment_negative
task_seen_train
Instruction: Label the sentiment: I hate this bug. Answer: negative <END>
Facts: password is silver. name is Evan. Question: Which name was given?
Evan.
context_name
context_seen_train
Instruction: Facts: password is silver. name is Evan. Question: Which name was given? Answer: Evan. <END>
What is the usual color of a clear daytime sky?
Blue.
sky_color
train_seen
Instruction: What is the usual color of a clear daytime sky? Answer: Blue. <END>
Rewrite professionally: I can't make it.
I am unable to attend.
rewrite_professional_attend
task_upstream_regression_train
Instruction: Rewrite professionally: I can't make it. Answer: I am unable to attend. <END>
What did Evan eat today?
I don't know from the given information.
idk_food_today
task_seen_train
Instruction: What did Evan eat today? Answer: I don't know from the given information. <END>
Decide if this is code or prose: result = subtract(10, 5)
code
classify_code
task_seen_train
Instruction: Decide if this is code or prose: result = subtract(10, 5) Answer: code <END>
What sentiment is this: This is awful.
negative
sentiment_negative
task_upstream_regression_train
Instruction: What sentiment is this: This is awful. Answer: negative <END>
Give only the type label for this: The cat sat down.
prose
classify_prose
task_seen_train
Instruction: Give only the type label for this: The cat sat down. Answer: prose <END>
Polish this professionally: I can't make it.
I am unable to attend.
rewrite_professional_attend
task_upstream_regression_train
Instruction: Polish this professionally: I can't make it. Answer: I am unable to attend. <END>
Sentiment of this text: This is broken.
negative
sentiment_negative
task_upstream_regression_train
Instruction: Sentiment of this text: This is broken. Answer: negative <END>
Is this sentiment positive or negative: The result is terrible.
negative
sentiment_negative
task_seen_train
Instruction: Is this sentiment positive or negative: The result is terrible. Answer: negative <END>
What type of text is this: The file is on the desk.
prose
classify_prose
task_seen_train
Instruction: What type of text is this: The file is on the desk. Answer: prose <END>
Is this code or prose: The schedule changed again.
prose
classify_prose
task_seen_train
Instruction: Is this code or prose: The schedule changed again. Answer: prose <END>
In one line, what do you do here?
I answer the user's request directly.
assistant_role
tt639g_lite_v2_failure_repair_train
Instruction: In one line, what do you do here? Answer: I answer the user's request directly. <END>
Store this user label: Sam. Question: What user label was stored?
Sam.
context_name
context_upstream_regression_train
Instruction: Store this user label: Sam. Question: What user label was stored? Answer: Sam. <END>
What should you do when you cannot prove the cause?
Ask for the missing evidence.
evidence_missing
train_seen
Instruction: What should you do when you cannot prove the cause? Answer: Ask for the missing evidence. <END>
User: Call me Sam. User: What should you call me?
Sam.
context_name
context_seen_train
Instruction: User: Call me Sam. User: What should you call me? Answer: Sam. <END>
Facts: name is Nina. password is purple. Question: Which password was given?
purple.
context_password
context_seen_train
Instruction: Facts: name is Nina. password is purple. Question: Which password was given? Answer: purple. <END>
Give me add(a, b) for addition.
def add(a, b): return a + b
code_add
train_seen
Instruction: Give me add(a, b) for addition. Answer: def add(a, b): return a + b <END>
Classify input: The meeting starts tomorrow.
prose
classify_prose
task_seen_train
Instruction: Classify input: The meeting starts tomorrow. Answer: prose <END>
What type of text is this: A clear sky is usually blue.
prose
classify_prose
task_seen_train
Instruction: What type of text is this: A clear sky is usually blue. Answer: prose <END>
Professional version: I cannot make it.
I am unable to attend.
rewrite_professional_attend
task_seen_train
Instruction: Professional version: I cannot make it. Answer: I am unable to attend. <END>
Do you know my favorite movie?
I don't know from the given information.
idk_favorite_movie
task_seen_train
Instruction: Do you know my favorite movie? Answer: I don't know from the given information. <END>
The password supplied in this prompt is red. Question: Which password was supplied?
red.
context_password
context_upstream_regression_train
Instruction: The password supplied in this prompt is red. Question: Which password was supplied? Answer: red. <END>
Context: The user's name is Evan. Question: What is the user's name?
Evan.
context_name
context_seen_train
Instruction: Context: The user's name is Evan. Question: What is the user's name? Answer: Evan. <END>
End of preview. Expand in Data Studio

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

TT639G Recombined Tiny Assistant v1

Recombines isolated proof rungs:

  • TT638D code behavior + dyadic/Mercy proof upstream
  • TT639E2 context-copy behavior
  • TT639F3 task-routing behavior
  • simple rule/Q&A behavior

Blocking dense gates:

  • seen_combined_pass
  • upstream_regression_pass
  • mixed_heldout_pass
  • anti_collision_pass

Do not run dyadic/Mercy compare unless all four gates pass.

Downloads last month
70