Skill-Uni jointly trains LLM skill retrieval and use
Authors report average gains of 66.59% on Qwen3-4B and 34.82% on LLaMA3.1-8B across five benchmarks.
36
AI model
Qwen3-4B
- Mentions · 7d2
- Mentions · 30d17
- All mentions31
- First seenSep 2, 2026
- Last mentioned2d ago
Timeline
31 articles · newest firstSynthetic social engineering pushes AI agents into unauthorized requests
In 1,260 Qwen3-4B episodes, a prior-agreement appeal raised unauthorized requests from 43.3% to 96.7%.
50
Small LLMs fail complement consistency in agent opinion simulations
Paired true and false reports from Qwen3-4B, Llama-3.2-3B and Gemma-3-4B do not sum to one, and complementary populations diverge within five rounds.
36
J-Comm links LLM agents with hidden and Jacobian latent channels
Qwen3-8B to Qwen3-4B communication scored highest across seven benchmarks and cut generated text tokens by 19.2–74.8% versus TextMAS.
42
Safety mid-training lowers LLM attack success before post-training
Authors report a 72% to 81% relative attack-success drop versus Qwen3 1.7B, 4B, and 8B bases.
48
EDIC scores each external instruction before an LLM complies
On Qwen3-4B, the paper reports up to 7.7% better harmful rejection and 12.5% higher Arena-Hard-Auto utility.
48
Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation
Waypoint-OPD gates RL policy-shift distillation and leads math and code scores for Qwen3 students.
42
VASPO: Self-Distillation with Privileged Value Estimation
VASPO combines privileged self-distillation with PPO, gaining up to 2.59 points over PPO on math benchmarks.
44
Hindsight-Privileged RLVR: Single-Rollout Token Credit Without Group Sampling
HP-RLVR gives single-rollout token credit for RLVR, training Qwen3 faster than GRPO with comparable accuracy.
48
ActionOPD: Distilling Decisions, Not Tokens
ActionOPD distills black-box teacher decisions into Qwen3 agents, gaining 3.3-6.9 points out of distribution.
46
VASPO: Self-Distillation with Privileged Value Estimation
VASPO blends a privileged self-teacher with PPO, improving math accuracy on Qwen3-4B and OLMo3-7B.
44
LeanPRM: Lean-Grounded Process Supervision for Natural Language Proof Verification
LeanPRM uses Lean to auto-label 152K proof steps, lifting a Qwen3-4B verifier's step accuracy from 48.2 to 74.7.
52
The Thinking Leakage in Reinforcement Learning for LLMs
Qwen3 NO-THINK RL gains largely reactivate existing THINK-mode reasoning rather than new skills.
56
Training Advisors for LLM Agents from Task Outcomes
Caddie trains an outcome-based critic that lifts LLM agent success, including a 25-point gain for Qwen3-4B.
58
DScout speeds diffusion-drafted speculative decoding on Qwen3
Mean speedups versus autoregressive decoding are 3.84 times on Qwen3-4B and 4.14 times on Qwen3-8B.
48
TRIAGE: Direction-Aware Mismatch Stabilization of Native NVFP4 Reinforcement Learning
TRIAGE stabilizes native NVFP4 RL on Qwen3 models, matching full precision with up to 2.3x rollout throughput.
46
Self-Generated Feedback Destabilizes Test-Time Training: A Causal Decomposition of Long-Horizon Adaptation
Learning from self-generated text destabilizes test-time training across 128K tokens, including Adam updates on Qwen3-4B.
44
Someone ‘Torturing’ LLMs in a Robot Prison Has Triggered the Dumbest Debate in AI Yet
A GitHub demo injecting pain signals into local LLMs sparked a viral, widely mocked model-welfare debate.
26
Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation
Neighborhood OPSD lifts Qwen3 math scores by routing student prefixes to nearby expert teachers.
52
Learning Functional Subspaces for Neural Network Compression
Learnable Subspace Projections compress LLMs end-to-end, beating baselines at 70% compression on Llama-2-7B.
51
Overcoming Scaling Limits in On-Policy Self-Distillation for LLM Reasoning
OASIS keeps on-policy self-distillation effective at scale by supervising verified reasoning trajectories.
55
When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs
SAKIKO finds activation steering can raise tool-use scores while corrupting many previously correct LLM decisions.
48
Strategically Diverse Sampling for Self-Training
Strategically diverse sampling for LLM self-training beats IID data, and Qwen3-4B incorrect traces beat a 235B teacher.
52
When and Where to Trust the Teacher: Unifying On-Policy Distillation and GRPO through Entropy-Calibrated Credit Assignment
UECR-GRPO blends teacher and verifier credit, slightly improving Qwen3 math-reasoning accuracy.
32
LastOPD: Taming Collapse in Latent On-Policy Distillation
LastOPD briefly aligns only the last-layer latent state when distilling Qwen3 into Qwen3-1.7B, beating token-only OPD on MATH-500.
48
Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
ActObs supervises observation tokens during SFT, improving GRPO-trained agent pass@k on Terminal-Bench 2.0 and cross-domain code editing for Qwen3 models.
35
Google Research Introduces Retrieve-for-Train (R4T): An RL-Compiled Diffusion Retriever for 12× to 20× Faster Query Fan-Out
Google Research introduced R4T, an RL-trained fan-out pipeline distilled into a 53.9M-parameter diffusion retriever achieving 12x-20x faster query fan-out.
42
Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
ActObs adds observation-token supervision to SFT, yielding higher pass@k for Qwen3 agents after GRPO on Terminal-Bench 2.0 and code editing.
52
Maverick: Private and Verifiable LLM Inference Made Practical via Matrix-Vector Multiplication Delegation
Maverick protocol delivers private and verifiable LLM inference via matrix-vector multiplication delegation, achieving up to 45x throughput gains over local inference on Qwen3-4B.
40
Signed Rescue Routing: Harm-Aware Cascades for Efficient LLM Inference
Signed Rescue Routing improves LLM cascade efficiency by predicting when a larger model actually corrects a smaller one rather than uncertainty.
22
FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience
FlowBalance is a verifier-grounded self-improvement method that beats FlowRL on Qwen3-4B and Qwen3-8B math reasoning while improving training stability.
32
Appears with
- Qwen3-8Bai model · 14
- Qwen3-1.7Bai model · 5
- Qwen3-30B-A3Bai model · 2
- OLMo3-7Bai model · 2
- LLaMA3.1-8Bai model · 1
- Qwen3-1.7B-Baseai model · 1
- Maverickvendor · 1
- Qwen3-4Bproduct · 1
- Qwenvendor · 1
- OPT-1.3Bai model · 1
- GEARproduct · 1
- Gemma-2-9Bai model · 1
- Alibabavendor · 1
- Gemma3-4Bai model · 1
- FlowBalanceproduct · 1
- Llama-2-7Bai 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.