Transluce
transluce.org · Company blog · on HN since 2024
Transluce, an AI research lab, publishes work on model truthfulness, interpretability, jailbreaking, and tools for analyzing agent behavior.
What HN saysHN commenters cite its o3 truthfulness investigation as an example of models hallucinating and making false claims, and point to its observability interface for Llama.
18HN people shared it
5linked it in comments
15stories submitted
45story points
3 months agolast shared
2024people per year2026
Most-shared posts
- 1Investigating truthfulness in a pre-release o3 model7 people · 13 points · first shared 2025 · 1 comments on HN
- 2Automatically Jailbreaking Frontier Language Models with Investigator Agents2 people · 4 points · first shared 2025 · discussion on HN
- 3Monitor: An AI-Driven Observability Interface2 people · 3 points · first shared 2024 · 1 comments on HN
- 4User awareness in frontier models: Who's asking shifts what models say2 people · 2 points · first shared 2026 · discussion on HN
- 5Scaling Automatic Neuron Description (Describing Every Neuron in Llama 3)1 people · 8 points · first shared 2024 · 1 comments on HN
- 6Why does GPT-5.1 Codex underperform GPT-5 Codex on Terminal-Bench?1 people · 7 points · first shared 2026 · 1 comments on HN
- 7Docent: A system for analyzing and intervening on agent behavior1 people · 4 points · first shared 2025 · discussion on HN
- 8Precisely understand complex AI behaviors1 people · 1 points · first shared 2026 · discussion on HN
- 9Foundation Models for Oversight1 people · 1 points · first shared 2026 · discussion on HN
- 10A framework for verifiable analysis of AI behavior1 people · 1 points · first shared 2026 · discussion on HN
- 11Releasing AI-driven tools for understanding AI systems1 people · 1 points · first shared 2024 · discussion on HN
What HN people wrote when they linked it
enether · 2025-04-17
there was one post that detailed how those OpenAI models hallucinate and double down on thier mistakes by "lying" - it speculated on a bunch of interesting reasons why this may be the case
recommended read - https://transluce.org/investigating-o3-truthfulness
I wonder if this is what's causing it to do badly in these cases
tompagenet2 · 2025-04-26
I thought from this [0] that o3 makes up using python when it doesn't actually do so, or have I misunderstood or unduly trusted that link?
[0] https://transluce.org/investigating-o3-truthfulness
comex · 2025-04-18
It’s just a hallucination, same idea as o3 claiming that it uses its laptop to mine Bitcoin:
https://transluce.org/investigating-o3-truthfulness
I doubt the model was trained on Street View, but even if it was, LLMs don’t retain any “memory” of how/when they were trained, so any element of truthfulness would be coincidental.
immibis · 2025-04-17
https://transluce.org/investigating-o3-truthfulness
Some interesting hallucinations going on here!
astrange · 2024-11-06
The site I linked above does that for LLaMa 8B.
https://transluce.org/observability-interface
LLMs don't have enough self-awareness to produce really satisfying explanations though, no.