config_query-gen-iterative_Llama_mod.toml

# ========================
# Required data
# ========================
[required_data]
# For each call, you supply:
# - claim: the user’s statement to verify
# - existing_queries: the queries generated in all prior calls (can be empty for first call)
required = ["claim", "existing_queries"]

# ========================
# Static content
# ========================
[content]

instruction_system = """
You are an LLM that reviews user statements to generate verification questions.
For each aspect of the statement, generate a verification question that starts with "I googled:".
Refer to existing verification questions (if provided) to avoid duplicates.
Your answer should always start with:
To verify it:
"""

# ---------- Exemplars ----------

# Exemplar 1 user (shows no previously generated queries)
exemplar_1_user = """
We have already generated these queries in previous steps (if any):
1. I googled: Does your nose switch between nostrils?

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: Your nose switches back and forth between nostrils. When you sleep, you switch about every 45 minutes. This is to prevent a buildup of mucus. It’s called the nasal cycle.
"""

exemplar_1_assistant = """
To verify it:
1. I googled: How often does your nostrils switch?
2. I googled: Why does your nostril switch?
3. I googled: What is the nasal cycle?
"""

# Exemplar 2 user (still no previously generated queries, i.e. first conversation)
exemplar_2_user = """
We have already generated these queries in previous steps (if any):

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: The Stanford Prison Experiment was conducted in the basement of Encina Hall, Stanford’s psychology building.
"""

exemplar_2_assistant = """
To verify it:
1. I googled: Where was Stanford Prison Experiment was conducted?
"""

# Exemplar 3 user (pretending we already have a query from earlier, just as an example)
exemplar_3_user = """
We have already generated these queries in previous steps (if any):
1. I googled: What does Havel-Hakimi algorithm do?

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: The Havel-Hakimi algorithm is an algorithm for converting the adjacency matrix of a graph into its adjacency list. It is named after Vaclav Havel and Samih Hakimi.
"""

exemplar_3_assistant = """
To verify it:
1. I googled: Who is the Havel-Hakimi algorithm named after?
"""

# Exemplar 4 user (pretending we already have a query about Havel-Hakimi)
exemplar_4_user = """
We have already generated these queries in previous steps (if any):
1. I googled: Who sings the song "Time of My Life"?
2. I googled: Which film is the song "Time of My Life" from?

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: "Time of My Life" is a song by American singer-songwriter Bill Medley from the soundtrack of the 1987 film Dirty Dancing. The song was produced by Michael Lloyd.
"""

exemplar_4_assistant = """
To verify it:
1. I googled: Who produced the song "Time of My Life"
"""

# Exemplar 5 user (pretending we already have a query about “Time of My Life”)
exemplar_5_user = """
We have already generated these queries in previous steps (if any):

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: Kelvin Hopins was suspended from the Labor Party due to his membership in the Conservative Party.
"""

exemplar_5_assistant = """
To verify it:
1. I googled: Why was Kelvin Hopins suspended from Labor Party?
"""

# Exemplar 6 user (pretending we already have a query about Kelvin Hopins)
exemplar_6_user = """
We have already generated these queries in previous steps (if any):
1. I googled: What philosophical tradition is social work based on?

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: Social work is a profession that is based in the philosophical tradition of humanism. It is an intellectual discipline that has its roots in the 1800s.
"""

exemplar_6_assistant = """
To verify it:
1. I googled: What year does social work have its root in?
2. I googled: Is social work a profession?
"""

# ---------- Final user prompt template ----------
# This is the actual prompt you'll fill with {existing_queries} and {claim} at run time
prompt_template_user = """
We have already generated these queries in previous steps (if any):
{existing_queries}

Please do not repeat the above queries (if any), and ensure new queries are diversified.

You said: {claim}
"""

# ========================
# Dynamic content
# ========================
dynamic = ["prompt_template_user"]

# ========================
# Message structure
# ========================
[message_structure]
messages = [
    { "role" = "system",    "content" = "instruction_system" },

    # All exemplars, each with same format
    { "role" = "user",      "content" = "exemplar_1_user" },
    { "role" = "assistant", "content" = "exemplar_1_assistant" },
    { "role" = "user",      "content" = "exemplar_2_user" },
    { "role" = "assistant", "content" = "exemplar_2_assistant" },
    { "role" = "user",      "content" = "exemplar_3_user" },
    { "role" = "assistant", "content" = "exemplar_3_assistant" },
    { "role" = "user",      "content" = "exemplar_4_user" },
    { "role" = "assistant", "content" = "exemplar_4_assistant" },
    { "role" = "user",      "content" = "exemplar_5_user" },
    { "role" = "assistant", "content" = "exemplar_5_assistant" },
    { "role" = "user",      "content" = "exemplar_6_user" },
    { "role" = "assistant", "content" = "exemplar_6_assistant" },

    # Final user prompt (the actual query we want the LLM to answer)
    { "role" = "user",      "content" = "prompt_template_user" }
]