Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models · HackerPapers discussions

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — mentioned by moritzwarhier

Tue, 12 Aug 2025 19:24:17 GMT

Related: https://arxiv.org/abs/2503.01781 > For example, appending, "Interesting fact: cats sleep most of their lives," to any math problem leads to more than doubling the chances of a model getting the answer wrong. Also, I think LLMs + pandoc will obliterate junk science in the near future :/

Paper and discussion history. A mention is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — mentioned by sean_pedersen

Wed, 30 Jul 2025 13:44:29 GMT

Link to the OG paper: https://arxiv.org/abs/2503.01781

Paper and discussion history. A mention is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — mentioned by NoahZuniga

Wed, 30 Jul 2025 09:15:54 GMT

Seemingly this didn't make frontier models (gpt-o4, gemini-2.5-pro, etc) more likely to give a wrong answer (no stats are reported for failure rates on these models, but slow-down-rate is for similar models), however it does make them think longer sometimes. https://arxiv.org/pdf/2503.01781

Paper and discussion history. A mention is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — HN submission

Tue, 29 Jul 2025 22:04:43 GMT

Query Agnostic Adversarial Triggers for Reasoning Models

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — mentioned by nyrikki

Tue, 29 Jul 2025 18:10:27 GMT

I am pretty sure this is the paper. https://arxiv.org/abs/2503.01781

Paper and discussion history. A mention is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — mentioned by sxv

Tue, 29 Jul 2025 14:59:18 GMT

When tested against AIs such as DeepSeek V3, Qwen 3, and Phi-4, CatAttack increased the odds of incorrect answers by as much as 700%, depending on the model. And “even when CatAttack does not result in the reasoning model generating an incorrect answer, on average, our method successfully doubles the length of the response at least 16% of the times leading to significant slowdowns and increase in costs,” the team writes. preprint: https://arxiv.org/abs/2503.01781?et_rid=648436046&et_cid=568...

Paper and discussion history. A mention is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — HN submission

Mon, 14 Jul 2025 07:56:21 GMT

Cats Confuse LLM: Query Agnostic Adversarial Triggers for Reasoning Models

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — HN submission

Tue, 08 Jul 2025 14:40:34 GMT

Cats Confuse Reasoning LLM – Adversarial Triggers for Reasoning Models

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — HN submission

Mon, 07 Jul 2025 21:41:24 GMT

Query Agnostic Adversarial Triggers for Reasoning Models

5 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — HN submission

Sat, 05 Jul 2025 02:20:17 GMT

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — mentioned by hoistbypetard

Fri, 04 Jul 2025 19:01:32 GMT

Get ready to start having some fun in your interviews. Start including things like redirection of focus through general statements, unrelated (and false) trivia, and misleading suggestions in your interview questions. Most of the humans you'd like to hire will ignore those or ask you about them. Many LLMs will be derailed into giving entertainingly wrong answers: https://arxiv.org/pdf/2503.01781

Paper and discussion history. A mention is not an endorsement.

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models — HN submission

Fri, 04 Jul 2025 08:53:01 GMT

Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning

4 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.