[BUG] Jump forward w/ outlines backend slightly changes the decoding results

[merrymercy]

Observation

For a json schema, outlines w/o jump forward and outlines w/ jump forward give slightly different results. We want to understand whether it is a bug or it is expected.

outlines w/ jumpforward
{ "name" : "Paris" , "population" : 2 }

outlines w/o jumpforward
{ "name": "Paris", "population": 2140000 }

xgrammar w/ jumpforward
{"name": "Paris", "population": 2140000}

xgrammar w/o jumpforward
{"name": "Paris", "population": 2140000}

Reproduce

python3 -m sglang.launch_server --model meta-llama/Meta-Llama-3.1-8B

import json
import requests

json_schema = json.dumps(
    {
        "type": "object",
        "properties": {
            "name": {"type": "string"},
            "population": {"type": "integer"},
        },
        "required": ["name", "population"],
    }
)

# JSON
response = requests.post(
    "http://localhost:30000/generate",
    json={
        "text": "Here is the information of the capital of France in the JSON format.\n",
        "sampling_params": {
            "temperature": 0,
            "max_new_tokens": 32,
            "json_schema": json_schema,
        },
    },
)
print(response.json())