agastyasridharan/olens-llama3.1-70b-base
AI model by agastyasridharan
Older snapshot: check the original source for current information.
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
This repo holds an activation verbalizer for the base model meta-llama/Llama-3.1-70B. It reads one residual-stream vector at one of 12 layers (28, 32, …, 72) and writes K phrases of N tokens each, one per line. The phrases are the ones whose activation signatures, superposed with non-negative weights, best reconstruct that vector. The method follows the Oracle Lens appendix of Anthropic's workspace paper (transformer-circuits.pub, 2026): This is the companion of agastyasridharan/olens-tulu3-70b, which is the same recipe on allenai/Llama-3.1-Tulu-3-70B, the instruct model fine-tuned from this base. The two lenses were trained on identical text: the same WildChat prompts, Tulu-3-70B's replies, and the same token positions. So they differ only in whose activations they read. Use the Tulu-3 tokenizer and chat template (allenai/Llama-3.1-Tulu-3-70B). The oracle was trained with them, on top of the base weights. The marker token (id 128108) carries the activation: its input embedding is replaced by 16000·h/‖h‖. The prompt…
Read the full model card ↗ · Preview checked 2026-10-28T14:03:57.745Z
Inside the original model card — Document outline
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Documentation belongs to its respective authors. Reported project/model license: llama3.1. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
agastyasridharan/olens-llama3.1-70b-base is a task-unspecified model repository published by agastyasridharan on Hugging Face. A library was not reported.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- Not reported
- Library
- Not reported
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- llama3.1
- Access
- Not gated by Hugging Face
- Created
- 2026-10-12T06:36:19.000Z
- Last modified
- 2026-10-12T06:37:37.000Z
Compatibility and lineage
Language coverage was not reported.
Reported base models: meta-llama/Llama-3.1-70B
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-10-12T06:41:18.150Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.