Llama-3.1-8B

Type: AI model

Llama-3.1-8B is an 8-billion-parameter open-weight model in Meta’s Llama 3.1 family, widely used for fine-tuning and on-device or self-hosted inference. It is distributed on Hugging Face and remains a common baseline in open-model research. GROUNDING tracks Llama-3.1-8B’s adoption, derivatives, and benchmark standing.

Recent Updates

• 2026-06-29: VASAE aligns SAE features with token vocabulary during training (cs.CL updates on arXiv.org) · arxiv.org — Embeddings GPT-2 small
• 2026-07-01: Guideline for customizing generative AI agents for transportation engineering (cs.AI updates on arXiv.org) · arxiv.org — LLM Evals Qwen2.5-7B
• 2026-07-02: FRAME proposes learnable adaptation domains for PEFT adapters (cs.LG updates on arXiv.org) · arxiv.org — arXiv · Qwen2.5-7B
• 2026-07-03: Mechanistic analysis of authority bias in LLM sycophancy (cs.CL updates on arXiv.org) · arxiv.org — Qwen3-8B · Gemma-2-9B
• 2026-07-03: Cybersecurity LLM specialization via minimal-token domain-adaptive pretraining (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Llama · DeepSeek · Qwen Ahmed Mohamed Hussain DeepSeek-R1-Distill-Qwen-14B · Llama-3.3-70B-Instruct Llama-Primus-Base Foundation-Sec-8B
• 2026-07-03: LLM robustness to science skepticism behaves in three different ways (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Qwen2.5-7B · Mistral-7B
• 2026-07-03: BPE tokenization creates safety gaps that bypass refusal behavior in several LLM families (cs.AI updates on arXiv.org) · arxiv.org — LLM Evals Qwen 3-4B · Qwen 2.5 7B · Gemma-3-4B · Mistral-7B
• 2026-07-03: AIriskEval-edu introduces a dataset for auditing K-12 instructional explanations (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Open Source LLMs
• 2026-07-03: Breaking Safety at the Token Boundary: BPE Tokenization and Alignment Gaps in LLMs (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Qwen 3-4B · Qwen 2.5 7B · Gemma-3-4B · Mistral-7B
• 2026-07-08: LongCrafter synthesizes long-context SFT data with evidence graphs (cs.CL updates on arXiv.org) · arxiv.org — Long Context LLM Evals arXiv · Qwen2.5-7B
• 2026-07-08: RL reward design materially changes LLM process-model generation quality (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Qwen 2.5-14B
• 2026-07-09: SmartHomeSecure uses constrained LLM repair for Home Assistant YAML errors (cs.AI updates on arXiv.org) · arxiv.org — LLM Evals home-assistant · gpt-oss-20b · GPT-OSS 120B · Llama-3.3-70b
• 2026-07-13: GrAInS: Gradient-based attribution for inference-time steering of LLMs and VLMs (cs.CL updates on arXiv.org) · arxiv.org — Duy Nguyen LLaVA-1.6-7B
• 2026-07-14: Token probability differences between production and perception prompts in LLMs (cs.CL updates on arXiv.org) · arxiv.org — arXiv · Llama EuroLLM-9B gemma-2-9b-it · Mistral-7B-Instruct-v0.3 · Qwen2.5-7B-Instruct
• 2026-07-14: Probing LLM internal states to detect confident hallucinations in financial QA (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Qwen3-8B · Gemma-2-9B
• 2026-07-15: MAGE: Stability–Performance Trade-offs in Multi-Component Prompt Optimization (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals gpt-4o-mini
• 2026-07-17: Branching Policy Optimization for sandbox-native language agent RL (cs.LG updates on arXiv.org) · arxiv.org — Agents LLM Evals Qwen2.5-7B
• 2026-07-20: VarRate proposes training-free variable-rate KV cache compression for long-context LLMs (cs.CL updates on arXiv.org) · arxiv.org — Long Context Context Engineering Qwen2.5-7B
• 2026-07-24: Benchmark Finds Large Language Models Miss Multi-Sensor Hazard Signals (cs.AI updates on arXiv.org) · arxiv.org — LLM Evals ChatGPT-4o Gemini 2.5 Flash · DeepSeek · Kimi
• 2026-07-24: Inference-time knowledge injection improves zero-shot delirium prediction in open-weight LLMs (cs.CL updates on arXiv.org) · arxiv.org — Context Engineering LLM Evals arXiv · Llama-3.3-70b · GPT-5.2
• 2026-07-24: Monkey King Bang: A Unified Scientific Multimodal Foundation Model (cs.LG updates on arXiv.org) · github.com — arXiv · Hugging Face · Meta sais-org MKB · Qwen3-VL · Qwen3-VL-8B ESM-2 ConvFormers Swin-ViT · SAM 3 Biology-Instructions Intern-S1-Pro BiomedParse HRES
• 2026-07-24: Pulsar Attention reduces distributed long-context inference cost with content-aware summaries (cs.CL updates on arXiv.org) · arxiv.org — Long Context LLM Evals
• 2026-07-24: Paper studies personality steering in LLMs with Jungian cognitive functions (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals arXiv
• 2026-07-28: Chain-of-Thought Unfaithfulness Detection Fails on Incorrect Answers (cs.CL updates on arXiv.org) · arxiv.org — LLM Evals Qwen 2.5 7B

FAQ

What is Llama-3.1-8B?

Llama-3.1-8B is an 8-billion-parameter open-weight model in Meta’s Llama 3.1 family, widely used for fine-tuning and on-device or self-hosted inference. It is distributed on Hugging Face and remains a common baseline in open-model research. GROUNDING tracks Llama-3.1-8B’s adoption, derivatives, and benchmark standing.

What does this page track?

Dated radar mentions, source links, related concepts, and builder-relevant context for Llama-3.1-8B, collected automatically by GROUNDING.

When was Llama-3.1-8B last mentioned?

Llama-3.1-8B was most recently mentioned in a radar update dated 2026-07-28.

Category: Text / Language Models