From d11650e0e469fda8a516f8f001c90bfb2178eea5 Mon Sep 17 00:00:00 2001 From: Tyrin Todd Date: Sat, 25 Apr 2026 20:24:44 -0700 Subject: [PATCH] server-inference push halueval_dialogue --- README.md | 3 +++ config.json | 56 ++++++++++++++++++++++++++++++++++++++++++++++++++++ program.json | 48 ++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 107 insertions(+) create mode 100644 config.json create mode 100644 program.json diff --git a/README.md b/README.md index e69de29..1388a48 100644 --- a/README.md +++ b/README.md @@ -0,0 +1,3 @@ +--- +{} +--- \ No newline at end of file diff --git a/config.json b/config.json new file mode 100644 index 0000000..ce97bf1 --- /dev/null +++ b/config.json @@ -0,0 +1,56 @@ +{ + "model": null, + "signature": { + "description": "Determine whether the dialogue response is hallucinated given the reference knowledge and prior turns.As you reason, note any uncertainties, ambiguities, and gaps \u2014 both in how the instructions apply to the task at hand and in whether you have the knowledge needed to solve it. Be honest in your reasoning when you are unsure about your answer.", + "properties": { + "knowledge": { + "__dspy_field_type": "input", + "desc": "Reference knowledge used to ground the dialogue response.", + "prefix": "Knowledge:", + "title": "Knowledge", + "type": "string" + }, + "dialogue_history": { + "__dspy_field_type": "input", + "desc": "Conversation history preceding the response.", + "prefix": "Dialogue History:", + "title": "Dialogue History", + "type": "string" + }, + "response": { + "__dspy_field_type": "input", + "desc": "Dialogue response to evaluate for hallucination.", + "prefix": "Response:", + "title": "Response", + "type": "string" + }, + "reasoning": { + "__dspy_field_type": "output", + "desc": "Step-by-step reasoning, including uncertainty, ambiguity, or uncertainty in your thought process when relevant.", + "prefix": "Reasoning:", + "title": "Reasoning", + "type": "string" + }, + "hallucination": { + "__dspy_field_type": "output", + "desc": "Final hallucination label: 'yes' if the response is hallucinated, otherwise 'no'.", + "enum": [ + "yes", + "no" + ], + "prefix": "Hallucination:", + "title": "Hallucination", + "type": "string" + } + }, + "required": [ + "knowledge", + "dialogue_history", + "response", + "reasoning", + "hallucination" + ], + "title": "HaluEvalDialogueJudgeSig", + "type": "object" + } +} \ No newline at end of file diff --git a/program.json b/program.json new file mode 100644 index 0000000..c9d1d28 --- /dev/null +++ b/program.json @@ -0,0 +1,48 @@ +{ + "traces": [], + "train": [], + "demos": [], + "signature": { + "instructions": "Determine whether the dialogue response is hallucinated given the reference knowledge and prior turns.As you reason, note any uncertainties, ambiguities, and gaps \u2014 both in how the instructions apply to the task at hand and in whether you have the knowledge needed to solve it. Be honest in your reasoning when you are unsure about your answer.", + "fields": [ + { + "prefix": "Knowledge:", + "description": "Reference knowledge used to ground the dialogue response." + }, + { + "prefix": "Dialogue History:", + "description": "Conversation history preceding the response." + }, + { + "prefix": "Response:", + "description": "Dialogue response to evaluate for hallucination." + }, + { + "prefix": "Reasoning:", + "description": "Step-by-step reasoning, including uncertainty, ambiguity, or uncertainty in your thought process when relevant." + }, + { + "prefix": "Hallucination:", + "description": "Final hallucination label: 'yes' if the response is hallucinated, otherwise 'no'." + } + ] + }, + "lm": { + "model": "together_ai/Qwen/Qwen2.5-7B-Instruct-Turbo", + "model_type": "chat", + "cache": true, + "num_retries": 3, + "finetuning_model": null, + "launch_kwargs": {}, + "train_kwargs": {}, + "temperature": null, + "max_tokens": null + }, + "metadata": { + "dependency_versions": { + "python": "3.11", + "dspy": "3.1.3", + "cloudpickle": "3.1" + } + } +} \ No newline at end of file