AI tools to monitor orthogonal probe with the model inspector

📎 Click on me to view all my attachmentsTheme:

ListAttBBC | v6.12 | 2026-09-10 | Andrew.human & Claude.ai

#	File	Size	Downloads	Info	DL
00	Backfor more AI tools to monitor py _orthogonal_probe_ with the _model_inspector_ - Claude.pdf [native]	2.3 MB	3	ℹ️	⬇️
Attachment ID 7373 Filename (display) Backfor more AI tools to monitor py _orthogonal_probe_ with the _model_inspector_ - Claude.pdf File extension pdf MIME type Size 2.3 MB (2432256 bytes) Downloads 3 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash 65db27320aa26ef6bf8730c8a0fab7eee7f309ba Approved Yes Filename (on disk) 7373_65db27320aa26ef6bf8730c8a0fab7eee7f309ba Storage directory 2 (/var/www/html/attachments2) File exists on disk? Yes (core)
Chip — October 11, 2026, 09:43:37 AM
01	inspect_Qwen_Qwen2.5-0.5B.json	27.2 KB	0	ℹ️	⬇️
Attachment ID 7374 Filename (display) inspect_Qwen_Qwen2.5-0.5B.json File extension json MIME type Size 27.2 KB (27886 bytes) Downloads 0 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash d38cf7d36dd1e355e20a3458ab126d07c9a5ab99 Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
02	inspect_Qwen_Qwen2.5-0.5B.md	16.3 KB	0	ℹ️	⬇️
Attachment ID 7375 Filename (display) inspect_Qwen_Qwen2.5-0.5B.md File extension md MIME type Size 16.3 KB (16730 bytes) Downloads 0 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash 61d372a026905dc6bde3278d29ac605f5422d357 Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
03	inspect_selftest-tiny-random.json	25.1 KB	0	ℹ️	⬇️
Attachment ID 7376 Filename (display) inspect_selftest-tiny-random.json File extension json MIME type Size 25.1 KB (25672 bytes) Downloads 0 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash bcabdb95896a3c82f4c2e169f19fbbf9654ae339 Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
04	inspect_selftest-tiny-random.md	21.5 KB	0	ℹ️	⬇️
Attachment ID 7377 Filename (display) inspect_selftest-tiny-random.md File extension md MIME type Size 21.5 KB (21987 bytes) Downloads 0 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash 437c23fff2bb213753f306d72f3f4d748b7cbf99 Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
05	model_inspector_and_orthogonal_probe_v1.1.0.zip	46.9 KB	1	ℹ️	⬇️
Attachment ID 7378 Filename (display) model_inspector_and_orthogonal_probe_v1.1.0.zip File extension zip MIME type Size 46.9 KB (47983 bytes) Downloads 1 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash d90d0fc4d0537875ea77a3e8b87c04372cf7c9fc Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
06	qwen_chat_v1.0.1.zip	3 KB	1	ℹ️	⬇️
Attachment ID 7379 Filename (display) qwen_chat_v1.0.1.zip File extension zip MIME type Size 3 KB (3029 bytes) Downloads 1 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash 1241f5c9961a1060b40f51ff8d9a61be1aea0dce Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
07	qwen_repeat_v1.0.0.zip	10.5 KB	0	ℹ️	⬇️
Attachment ID 7380 Filename (display) qwen_repeat_v1.0.0.zip File extension zip MIME type Size 10.5 KB (10722 bytes) Downloads 0 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash 7fc2237a7eecdc6e2fb5a1a8b26faeef84863f4e Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.
08	AI tools to monitor py 'orthogonal_probe' with the 'model_inspector' -- Sysplex_Checkpoint_20261011_1851.md	11.7 KB	0	ℹ️	⬇️
Attachment ID 7397 Filename (display) AI tools to monitor py 'orthogonal_probe' with the 'model_inspector' -- Sysplex_Checkpoint_20261011_1851.md File extension md MIME type Size 11.7 KB (11962 bytes) Downloads 0 Dimensions n/a Uploaded by Guest Message ID 49761 Attachment type Normal attachment (0) Storage folder ID 2 Thumbnail attach ID none File hash 219dc6dc3d047a0bff502ecbeb42ab75c4a4bed8 Approved Yes Filename (on disk) not found on disk Storage directory unresolved File exists on disk? No
No recorded downloads yet.

dopetalk

News:

dopetalk does not endorse any advertised product nor does it accept any liability for it's use or misuse

Author Topic: AI tools to monitor orthogonal probe with the model inspector (Read 35 times)

Chip (OP)

Model inspection report: Qwen/Qwen2.5-0.5B

5. What the model does with your prompt

Run A (reference) - the prompt alone

Run B (test) - the prompt alone, with the memory 'direction' added to the running thought at layer 12, strength 0.5 x the typical size

Reference versus test: what did the push change?

How the change travels through the layers

Size of the running thought at every position (run B)

8. Self-checks (the report checking its own arithmetic)

Model inspection report: selftest-tiny-random

1. Identity and provenance

Model class and generation defaults

Tokenizer

Publisher's model card (VERBATIM - not derived from the weights)

This machine and software

2. Parameter census

Per layer

3. Attention anatomy

Shapes inside one layer (layer 0)

Memory the model needs per word-piece of context (the 'KV cache')

Attention versus MLP

4. Weights

Weight size per layer (Frobenius norm = overall magnitude of each matrix)

Fingerprint of all weights

5. What the model does with your prompt

Run A (reference) - the prompt alone

Run B (test) - the memory pasted as plain text in front of the prompt

Reference versus test: what did the push change?

How the change travels through the layers

Size of the running thought at every position (run B)

8. Self-checks (the report checking its own arithmetic)

Sysplex Checkpoint — 2026-10-11 18:51 AEDT (Clone orchestrating; AI tools chapter)

1. The three boxes

2. Done (carried over)

3. Open items (carried over; parked while the AI chapter ran — "forget the script")

4. This chapter: AI tools on Clone (Qwen2.5-0.5B)

4.1 Delivered (all Python 3.8 / torch 1.10 compatible)

4.2 What Clone has actually shown

4.3 Technical facts learned (keep for the next build)

4.4 The two-LLM "watcher" design — saved, NOT built, ABORTED 2026-10-11 12:20 (hardware)

4.5 Pending / not built

5. Andrew's stated next steps (his words)

6. Technical notes (carried over)

Need help or a chat ?

If you need any help or a chat then IM/PM or email me, Chip

dopetalk does not endorse any advertised product nor does it accept any liability for it's use or misuse