slim-nli

SLIM-NLI

<!-- Provide a quick summary of what the model is/does. -->

slim-nli is part of the SLIM ("Structured Language Instruction Model") model series, consisting of small, specialized decoder-based models, fine-tuned for function-calling.

slim-nli has been fine-tuned for natural language inference (nli) function calls, generating output consisting of a python dictionary corresponding to specified keys, e.g.:

    {"evidence": ["contradicts"]}

SLIM models are designed to generate structured outputs that can be used programmatically as part of a multi-step, multi-model LLM-based automation workflow.

Each slim model has a 'quantized tool' version, e.g., 'slim-nli-tool'.

Prompt format:

function = "classify"
params = "nli"
prompt = "<human> " + {text} + "\n" +
                      "<{function}> " + {params} + "</{function}>" + "\n<bot>:"

<details> <summary>Transformers Script </summary>

model = AutoModelForCausalLM.from_pretrained("llmware/slim-nli")
tokenizer = AutoTokenizer.from_pretrained("llmware/slim-nli")

function = "classify"
params = "evidence"

# expects two statements - the first is evidence, and the second is a conclusion  

text1 = "The stock market declined yesterday as investors worried increasingly about the slowing economy."  
text2 = "Investors are positive about the market."  

# the two statements are concatenated with optional/helpful "Evidence: " and "Conclusion: " added  

text = "Evidence: " + text1 + "\n" + "Conclusion: " + text2  

prompt = "<human>: " + text + "\n" + f"<{function}> {params} </{function}>\n<bot>:"

inputs = tokenizer(prompt, return_tensors="pt")
start_of_input = len(inputs.input_ids[0])

outputs = model.generate(
    inputs.input_ids.to('cpu'),
    eos_token_id=tokenizer.eos_token_id,
    pad_token_id=tokenizer.eos_token_id,
    do_sample=True,
    temperature=0.3,
    max_new_tokens=100
)

output_only = tokenizer.decode(outputs[0][start_of_input:], skip_special_tokens=True)

print("output only: ", output_only)  

# here's the fun part
try:
    output_only = ast.literal_eval(llm_string_output)
    print("success - converted to python dictionary automatically")
except:
    print("fail - could not convert to python dictionary automatically - ", llm_string_output)

</details> <details> <summary>Using as Function Call in LLMWare</summary>

from llmware.models import ModelCatalog
slim_model = ModelCatalog().load_model("llmware/slim-nli")

# input text - expects two statements - the first is evidence, and the second is a conclusion    
text1 = "The stock market declined yesterday as investors worried increasingly about the slowing economy."    
text2 = "Investors are positive about the market."  
text = "Evidence: " + text1 + "\n" + "Conclusion: " + text2  

response = slim_model.function_call(text,params=["evidence"], function="classify")

print("llmware - llm_response: ", response)

</details>

Model Card Contact

Darren Oberst & llmware team

Join us on Discord