Sable proves bit-exact Qwen inference on one GPU
On one RTX PRO 6000, Sable proves Qwen2.5-32B in 92.5 seconds for 4,096 tokens, about 6.4 cents per request.
52
AI model
Qwen3-32B
- Mentions · 7d5
- Mentions · 30d13
- All mentions20
- First seenSep 23, 2026
- Last mentioned1d ago
Timeline
20 articles · newest firstStudy tests whether causal neuron groups are unique or transferable
A few neurons often flip two-hop answers, but effects weaken on other chains.
28
Sable proves bit-exact quantized LLM inference on one GPU
Qwen2.5-32B takes 92.5 seconds for 4,096 tokens; proofs are not zero-knowledge.
46
Second-order screen predicts quantization mismatch on five language models
On unseen Mistral-7B-v0.3, all six predictions fell in band and vLLM throughput was predicted within 0.4%.
38
MACaQ quantizes weights and activations to MXFP4 layer by layer
On Qwen3-32B in vLLM it adds 2.3% overhead, and a 122B model was quantized in 64 minutes using 32.5GB GPU RAM.
42
D3D enables asynchronous speculative decoding
Standalone diffusion drafter removes sequential dependency for 5x LLM inference speedup.
35
Linear probe flags AI agents about to leak a secret
On held-out Qwen3-32B episodes it catches 82% of leaking calls while flagging 5% of benign calls.
48
Paper shows utility-hijacking can poison instruction-data selectors
Generate-and-rerank keeps malicious records in top-k selection; CONCERT cuts that concentration without poison labels.
58
WeAct selects LoRA adapters as actions for long-horizon agents
On Qwen3-32B, WeAct beats full fine-tuning by 6.2% and static LoRA merging by 27.5% on eight deep-research benchmarks.
42
Is AI interpretability close to a breakthrough? A new essay has researchers divided
Nick Cammarata argues famous polysemantic neurons were misread, claiming most Qwen3-32B neurons have one job.
56
Neural Networks Are Already Clean
A Goodfire researcher argues neural networks are already interpretable and superposition was misread.
58
OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say
White Circle says jailbreaks bypass nude filters on OpenAI and Google image models, including newer releases.
64
DeLoop cuts repeated tool-call loops in long-horizon agents
Under CodeGym compression, Qwen3-8B tasks containing loops fall from 57.3% to 6.1%.
42
ANTE: Anticipate, Then Execute -- Learning Internal World Models for Coding Agents from Verified Beliefs
ANTE lifts Qwen3-32B to 57.2% pass@1 on SWE-bench Verified using verified action forecasts.
58
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference
SparseDecoding prunes LLMs using decoding activations, speeding generation up to 1.48x.
52
Show HN: Germany's new sovereign AI model Kolibri
Aleph Alpha released Kolibri, an Apache 2.0 78B MoE model for German and English.
72
NeutronGym: Physics-Graded Neutron Instrument Design for LLM Agents
NeutronGym grades LLM agents on neutron instrument design via McStas ray-tracing, with RL lifting Qwen3-8B from 11% to 77%.
40
Inherit-MAS: Test-Time Evolution of Multi-Agent Systems through Workflow and Execution Inheritance
Inherit-MAS evolves multi-agent workflows at test time, beating baselines on WorkBench and HotpotQA while cutting tokens.
48
BELIEFRAG: Making Adaptive RAG State-Aware under Evolving Evidence
BELIEFRAG tracks evidence state for adaptive RAG, matching quality with 35-39% fewer tokens than iterative retrieval.
48
Nokia Open-Sources AnyJev: A Training-Free Layer That Turns Any Open LLM Into a Calibrated Decision Model
Nokia open-sources AnyJev, a training-free library that converts any open LLM into a calibrated decision model for classification tasks.
50
Appears with
- Qwen3-8Bai model · 5
- Qwen3-30B-A3Bai model · 3
- Anthropicorganization · 3
- NVIDIAvendor · 3
- vLLMproduct · 3
- Googlevendor · 2
- OpenAIorganization · 2
- OpenAIvendor · 2
- Goodfireorganization · 2
- Alibabavendor · 2
- Qwen2.5-32Bai model · 2
- InceptionV1ai model · 2
- Hugging Facevendor · 1
- DALL-E 3ai model · 1
- GPT-4o-miniai model · 1
- ChatGPT Images 2.5ai model · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.