Update docs

docs/evaluations.md

@@ -10,7 +10,7 @@ Tests may be run with the `eval` command followed by the test name, e.g., for th
 ontogpt eval EvalCTD 
 ```
 
-By default, the evaluation will only be executed over a subset of the test input.
+By default, the evaluation will only be executed over a subset of the test corpus.
 
 The exact number of inputs to run the test over can be controlled with the `--num-tests` option, like this:
 
@@ -20,6 +20,10 @@ ontogpt eval --num-tests 1 EvalCTD
 
 To run the full set of tests, set `num-tests` to the input count for a given evaluation, as defined below.
 
+For each document, the evaluation process will attempt to process the full text without preprocessing.
+
+If the `--chunking` option is used, then the input text will instead be chunked into segments of a few tokens each, essentially creating new queries for each segment. This alternate strategy may impact test results and will result in longer run time.
+
 ## BC5CDR
 
 *Test Name:* EvalCTD

docs/functions.md

@@ -388,6 +388,7 @@ See the Evaluations section for more details.
 Options:
 
 * `--num-tests INTEGER` - number of test iterations to cycle through. Defaults to 5.
+* `--chunking` / `--no-chunking` - If set, chunk input text, then prepare a separate prompt for each chunk. Otherwise the full input text is passed. Defaults to False.
 
 Example: