Model techniques map
Techniquespost-trainingreward modelling

implementation detail · filed under post-training

Per-token tool-error reward shaping

Assigns shaping feedback at the failing token or step to improve credit assignment for malformed tool calls.

source
1
model
1
lab adopt it
1
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

used 1

Documented in

Evidence

1 span quoted from the sources, strongest treatment first.

This per-token shaping focuses credit assignment on the failing step and discourages long chains of malformed tool calls before falling back on the terminal verifier.

usedpost trainingin Laguna XS.2Poolside

Filed alongside

Other methods under post-training :: reward modelling.