Ikibazo gihuza ibibazo · Hagati Urwego
Logit Lens hamwe na Interineti Hagati ya Decoding Ikibazo
Gerageza hagati yigihe gisigaye gisomwa, gisanzwe no kudahuza, guhuza affine probe, ibisobanuro byamagambo, softmax, nimbibi zo gusobanura impamvu.
Ikibazo 1 Bya 8
What readout does a plain logit lens reuse to obtain vocabulary scores from an intermediate state?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- What readout does a plain logit lens reuse to obtain vocabulary scores from an intermediate state?
- A required new image classifier for every layer
- The model's appropriate final normalization and unembedding output projection
- A random dictionary sort with no model parameters
- A separate retrained tokenizer for each prompt
- Why can the same output projection be applied mathematically at multiple residual-stream depths?
- Every layer must contain an identical copy of the final state
- The entire transformer is a single linear mapping
- All intermediate token scores are guaranteed identical
- The states have compatible hidden dimensions, although their representations may not be equally calibrated for that readout
- Ni ikihe kibazo 'cyateguwe neza' gikosora ugereranije n'inzira zisanzwe?
- Ikosora per-layer idahuye hamwe na affine yize ihinduka
- Ikuraho ibikenewe byose bitarimo gushyirwaho
- Bituma icyitegererezo cyihuta mugusuzuma
- Isimbuza softmax na argmax
- If an intermediate lens readout favors a correct factual token, what has been directly observed?
- The unique causal location of the model's factual knowledge
- A guarantee that all later output will be correct
- That the particular probe maps that state to a high score for the token
- The exact reasoning process of a conscious system
- Niki lens ya logit ifite akamaro kanini cyane?
- Kugabanya uburemere bw'icyitegererezo
- Guhindura indimi
- Guhishira ibikorwa by'icyitegererezo
- Kureba uburyo n'aho ubuhanuzi bukora mubwimbitse
- Why can a plain final-head readout be misleading on early transformer states?
- Early states always use a different vocabulary file
- Their representation distributions can differ from the final state for which the head was trained
- Every early-layer vector is necessarily unstructured noise
- No matrix operation can be applied before the final layer
- A lens uses a model with a fifty-thousand-entry vocabulary and a 768-dimensional hidden state. How many vocabulary logits does its unembedding produce?
- 768, one score per hidden coordinate only
- One, because the token must be chosen before projection
- Fifty thousand, one score per vocabulary entry
- Exactly the number of transformer layers
- What does softmax add to a logit-lens vocabulary-score display?
- A normalized distribution over the projected vocabulary scores
- Proof that every displayed probability is a calibrated belief
- A causal intervention on the model's hidden states
- A replacement for the output projection's vocabulary dimension