HFEPX · Eval paper review
Mengya Hu, Qiong Wei, Sandeep Atluri
Published
Apr 28, 2026
Citations
0
Trust level
Provisional
Usefulness score
Unavailable
Extraction confidence
0% (Provisional)
Derived from abstract and metadata only.
Signals refreshed
May 4, 2026
Signal extraction is still processing. This page currently shows metadata-first guidance until structured protocol fields are ready.
This page is a lightweight research summary built from the abstract and metadata while deeper extraction catches up.
All signals on this page are inferred from the abstract only and may be inaccurate. Do not use this page as a primary protocol reference.
Best use
Background context only
Use if you need
A provisional background reference while structured extraction finishes.
What to verify
Read the full paper before copying any benchmark, metric, or protocol choices.
Main weakness
This page is still relying on abstract and metadata signals, not a fuller protocol read.
Eval-fit score is unavailable until extraction completes.
If you are doing eval pipeline work, start here
Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification. While useful, these can hide how risk changes between a user's input and the model's response. We present a paired, transition-based analysis over 1250 prompt-response records with human-provided labels over four harm categories (Hate, Sexual, Violence, Self-harm) and ordinal severity levels aligned with the Azure AI Content Safety taxonomy. 61% of responses de-escalate harm relative to the prompt, 36% preserve the same severity, and 3% escalate to higher harm. A per-category persistence/drift-up decomposition identifies Sexual content as 3x harder to de-escalate than Hate or Violence, driven by persistence on already-sexual prompts, not by newly introducing sexual harm from benign inputs. Jointly measuring response relevance reveals an empirical signature of the helpfulness-harmlessness tradeoff: all compliance-escalation cases (from non-zero prompts) are relevance-3 (high-quality, on-task content at elevated severity), while medium-severity responses show the lowest relevance (64%), driven by tangential elaborations in Violence and Sexual categories.
These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.
None explicit
No explicit feedback protocol extracted.
"Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification."
None explicit
Validate eval design from full paper text.
"Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification."
Not reported
No explicit QC controls found.
"Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification."
Not extracted
No benchmark anchors detected.
"Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification."
Not extracted
No metric anchors detected.
"Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification."
Unknown
Rater source not explicitly reported.
"Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification."
This page is using abstract-level cues only right now. Treat the signals below as provisional.
Evaluation fields are inferred from the abstract only.
Safety evaluations of large language models (LLMs) typically report binary outcomes such as attack success rate, refusal rate, or harmful/not-harmful response classification.
Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.
