ML News
new
|
past
|
best
|
rss
|
submit
about
1.
Hybrid Reasoning Agents: LLM agents that think in latent space, act through text
(
arxiv.org
)
520 points
by
allinonai
23 days ago
|
hide
|
past
|
pdf
|
239 comments
2.
The Little Book of Generative AI Foundations: An Intuitive Mathematical Primer
(
arxiv.org
)
395 points
by
ibobev
7 days ago
|
hide
|
past
|
pdf
|
60 comments
3.
AI models can provably hide arbitrary capabilities
(
arxiv.org
)
282 points
by
sbulaev
29 days ago
|
hide
|
past
|
pdf
|
145 comments
4.
Why is AI generated language so hard to understand?
(
arxiv.org
)
259 points
by
canjobear
18 days ago
|
hide
|
past
|
pdf
|
205 comments
5.
The alien space of science: Coherent but cognitively unavailable directions
(
arxiv.org
)
214 points
by
martincsweiss
20 days ago
|
hide
|
past
|
pdf
|
89 comments
6.
Attention Is Not What You Need (2025)
(
arxiv.org
)
197 points
by
westurner
16 days ago
|
hide
|
past
|
pdf
|
72 comments
7.
Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation
(
arxiv.org
)
180 points
by
theanonymousone
2 days ago
|
hide
|
past
|
pdf
|
45 comments
8.
Proving and improving the Platonic Representation Hypothesis with JEPAs
(
arxiv.org
)
172 points
by
Anon84
10 days ago
|
hide
|
past
|
pdf
|
85 comments
9.
Memorilla: Latent Semantic Memory for LLMs
(
arxiv.org
)
125 points
by
handfuloflight
5 days ago
|
hide
|
past
|
pdf
|
13 comments
10.
Agentic backdoors via pretraining poisoning
(
arxiv.org
)
109 points
by
gmays
9 days ago
|
hide
|
past
|
pdf
|
31 comments
11.
USDB: Scaling Storage for Next-Generation Recommender Systems at ByteDance
(
arxiv.org
)
49 points
by
1equalsequals1
8 days ago
|
hide
|
past
|
pdf
|
8 comments
12.
The Surprising Effectiveness of Deleting Weights in LLM Post-Training
(
arxiv.org
)
29 points
by
theanonymousone
20 hours ago
|
hide
|
past
|
pdf
|
discuss
13.
When Modernization Breaks Working Code: A Study of LLMs on Legacy PHP
(
arxiv.org
)
22 points
by
tambourine_man
9 days ago
|
hide
|
past
|
pdf
|
14 comments
14.
Self-Evolving Agent Harness
(
arxiv.org
)
18 points
by
axis_official
26 days ago
|
hide
|
past
|
pdf
|
1 comment
15.
The Illusion of Thinking: Models Don't Actually Reason
(
arxiv.org
)
14 points
by
CGMthrowaway
12 days ago
|
hide
|
past
|
pdf
|
1 comment
16.
Multi-Token Prediction Done Right
(
arxiv.org
)
13 points
by
matt_d
7 days ago
|
hide
|
past
|
pdf
|
4 comments
17.
Compaction Laundering: Provenance Destruction in Context-Compacting Coding Agent
(
arxiv.org
)
9 points
by
nahsra
16 days ago
|
hide
|
past
|
pdf
|
2 comments
18.
The Thinking Leakage in Reinforcement Learning for LLMs
(
arxiv.org
)
8 points
by
ilreb
26 days ago
|
hide
|
past
|
pdf
|
discuss
19.
DeepSeek-v4.1-Flash: Pushing the Limits of KV Cache Compression
(
arxiv.org
)
6 points
by
jonbaer
14 days ago
|
hide
|
past
|
pdf
|
discuss
20.
Learning What to Investigate Next: Meta-Reasoning for Long-Horizon Agents
(
arxiv.org
)
6 points
by
matt_d
28 days ago
|
hide
|
past
|
pdf
|
discuss
21.
Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories
(
arxiv.org
)
5 points
by
olalonde
11 days ago
|
hide
|
past
|
pdf
|
1 comment
22.
Artificial Epanorthosis: Why LLMs overuse a classical rhetorical figure
(
arxiv.org
)
5 points
by
Anon84
6 days ago
|
hide
|
past
|
pdf
|
discuss
23.
A team of Gemini 4 Argon agents beats Minecraft
(
arxiv.org
)
5 points
by
burrito_brain
26 days ago
|
hide
|
past
|
pdf
|
discuss
24.
Fork-and-Flush: Escaping Idea Basins in Autoresearch Agents
(
arxiv.org
)
5 points
by
matt_d
29 days ago
|
hide
|
past
|
pdf
|
discuss
25.
LLMs tested on historical prediction markets, NONE outperformed the market
(
arxiv.org
)
4 points
by
dataMike
7 days ago
|
hide
|
past
|
pdf
|
2 comments
26.
278M PII detection model bests OpenAI, AWS and Microsoft
(
arxiv.org
)
4 points
by
skjhn
8 days ago
|
hide
|
past
|
pdf
|
2 comments
27.
Is Complexity an Illusion? (2024)
(
arxiv.org
)
4 points
by
the-mitr
21 days ago
|
hide
|
past
|
pdf
|
1 comment
28.
A Fix for Softmax and Cross-Entropy
(
arxiv.org
)
4 points
by
theanonymousone
14 days ago
|
hide
|
past
|
pdf
|
discuss
29.
Parallel Language Models via Variational Decoding
(
arxiv.org
)
4 points
by
E-Reverance
22 days ago
|
hide
|
past
|
pdf
|
discuss
30.
LLM Agents Can Easily Tamper with Their Own Traces
(
arxiv.org
)
4 points
by
Kaedon
29 days ago
|
hide
|
past
|
pdf
|
discuss
More
About
|
RSS
|
RSS (all)
|
HN arXiv