rephrasing & neutral language

[baileyb0t]

Tarak mentioned LLM responses in the CLEAN work that, while technically summarizing the police narrative, repeat and reinforce the police's version of events and view of the person who was the subject of police violence.

My hunch is that filtering these responses through an LLM call with examples included in the prompt could go a long way towards improving the factual accuracy of the summary while being sensitive to the nature of the report.

I probably won't have time to look into this for a bit but I think the llm-unpack task in US-II-HT has the right structure and expects I/O in the form of raw LLM response objects

[ayyubibrahimi]

Yes! Think we can also apply this concept to redaction