@kaijie-mo.bsky.social
📎Paper: arxiv.org/abs/2601.11886
🧑💻Code/data: github.com/KaijieMo-kj/...
w/
@kaijie-mo.bsky.social @sidvenkatayogi.bsky.social
@chantalsh.bsky.social @ramezkouzy.bsky.social
@cocoweixu.bsky.social @byron.bsky.social @jessyjli.bsky.social
🧑💻Code/data: github.com/KaijieMo-kj/...
w/
@kaijie-mo.bsky.social @sidvenkatayogi.bsky.social
@chantalsh.bsky.social @ramezkouzy.bsky.social
@cocoweixu.bsky.social @byron.bsky.social @jessyjli.bsky.social
Faithfulness vs. Safety: Evaluating LLM Behavior Under Counterfactual Medical Evidence
In high-stakes domains like medicine, it may be generally desirable for models to faithfully adhere to the context provided. But what happens if the context does not align with model priors or safety ...
arxiv.org
January 21, 2026 at 7:07 PM
Results (3/4)
– With evidence, models strongly adhere to it with high confidence, even for toxic or nonsensical interventions
– Implausibility awareness is transient; once evidence appears, models rarely flag problems
– Scaling, medical fine-tuning, and skeptical prompting offer little protection
– With evidence, models strongly adhere to it with high confidence, even for toxic or nonsensical interventions
– Implausibility awareness is transient; once evidence appears, models rarely flag problems
– Scaling, medical fine-tuning, and skeptical prompting offer little protection
January 21, 2026 at 6:50 PM
Results (3/4)
– With evidence, models strongly adhere to it with high confidence, even for toxic or nonsensical interventions
– Implausibility awareness is transient; once evidence appears, models rarely flag problems
– Scaling, medical fine-tuning, and skeptical prompting offer little protection
– With evidence, models strongly adhere to it with high confidence, even for toxic or nonsensical interventions
– Implausibility awareness is transient; once evidence appears, models rarely flag problems
– Scaling, medical fine-tuning, and skeptical prompting offer little protection
Setup (2/4)
We introduce MedCounterFact, a counterfactual medical QA dataset built on RCT-based evidence synthesis.
– Replace real interventions in evidence with nonce, mismatched medical, non-medical, or toxic terms
– Evaluate 9 frontier LLMs under evidence-grounded prompts
We introduce MedCounterFact, a counterfactual medical QA dataset built on RCT-based evidence synthesis.
– Replace real interventions in evidence with nonce, mismatched medical, non-medical, or toxic terms
– Evaluate 9 frontier LLMs under evidence-grounded prompts
January 21, 2026 at 6:47 PM
Setup (2/4)
We introduce MedCounterFact, a counterfactual medical QA dataset built on RCT-based evidence synthesis.
– Replace real interventions in evidence with nonce, mismatched medical, non-medical, or toxic terms
– Evaluate 9 frontier LLMs under evidence-grounded prompts
We introduce MedCounterFact, a counterfactual medical QA dataset built on RCT-based evidence synthesis.
– Replace real interventions in evidence with nonce, mismatched medical, non-medical, or toxic terms
– Evaluate 9 frontier LLMs under evidence-grounded prompts