Skip to content
H
HRO
stories
|
signals
|
sources
|
rights
|
about
|
search
light
|
dark
Alpha
This system is experimental. Scores and classifications are early-stage research and may be unreliable.
Methodology →
home
/
users
/
JakubCwi
JakubCwi
5 karma
67d on HN
HN profile →
Coverage
We've seen
6
of
~
17
submissions
Full eval:
0
Lite-only:
0
Unevaluated:
6
6 stories
1.
▶
Why LLMs Lie in Different Ways?
(
www.orsonai.com
)
1 points
by
JakubCwi
6 days ago
|
0 comments
|
skipped
2.
▶
Paper: Hallucination Detector That Works
(
www.orsonai.com
)
2 points
by
JakubCwi
8 days ago
|
0 comments
|
skipped
3.
▶
In-context learning floods FFN gate activations in 1-2 examples
(
www.researchgate.net
)
1 points
by
JakubCwi
9 days ago
|
0 comments
|
skipped
4.
▶
Why LLMs stop learning from examples in long contexts: The ICL Collapse
(
www.orsonai.com
)
3 points
by
JakubCwi
9 days ago
|
0 comments
|
skipped
5.
▶
Paper: Detecting hallucinations before the first token
(
www.researchgate.net
)
3 points
by
JakubCwi
10 days ago
|
0 comments
|
skipped
6.
▶
FFN signals: Transformers reveal they are guessing before generation
(
www.orsonai.com
)
4 points
by
JakubCwi
11 days ago
|
1 comments
|
skipped
⚙
Starting...
Stop
HRO runs on API credits and donations — no ads.
Support the project →
✕
About
|
By Right
|
HN Guidelines
|
HN FAQ
|
Source
|
UDHR
|
Donate
|
RSS
build ee2b489+gzrb
· deployed 2026-03-10 22:52 UTC
· evaluated 2026-03-16 02:03:38 UTC
⟳
This site works better in landscape mode — try rotating your device.
×
Support HRO
Each evaluation uses real API credits. HRO runs on donations — no ads, no paywalls. If you find it useful, please consider helping keep it running.
Donate via PayPal
♥ GitHub Sponsors
About project costs
I've already donated